Welcome! » Log In » Create A New Profile

Benchmarking Maglib Spindles?

Posted by Anonymous 
Benchmarking Maglib Spindles?
February 01, 2012 11:50AM
What is the recommended way to benchmark various RAID configs for maglib volumes please?

Environment will be dedupe with auxcopy to tape.

I know I can use IOMeter, but I'm interested in any specific Commvault tools and what kind of profile (IO size, type, distribution etc.) I should use.

Thanks,
Paul
--
MIRA Ltd

Watling Street, Nuneaton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registration GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

__._,_.___
Reply to sender (paul.hutchings < at > mira.co.uk?subject=Re%3A%20Benchmarking%20Maglib%20Spindles%3F) | Reply to group (commvault < at > yahoogroups.com?subject=Re%3A%20Benchmarking%20Maglib%20Spindles%3F) | Reply via web post | Start a New Topic
Messages in this topic (1)
Recent Activity:Visit Your Group

Commvault Documentation here:
http://documentation.commvault.com/
Text-Only (commvault-traditional < at > yahoogroups.com?subject=Change Delivery Format: Traditional), Daily Digest (commvault-digest < at > yahoogroups.com?subject=Email Delivery: Digest) • Unsubscribe (commvault-unsubscribe < at > yahoogroups.com?subject=Unsubscribe) • Terms of Use

.


__,_._,___
Benchmarking Maglib Spindles?
February 01, 2012 12:10PM
Hi Paul,

You can use CVDISKPERF and the SIDB2 tools. CVDISKPERF will help you benchmark your maglib while the SIDB2 tool will help you benchmark the disk(s) on which you will put the DDB. By the way, disk performance for the DDB is a lot more critical than the one for the maglib itself. The faster your DDB disk(s) the faster your dedup will run.

For the docs, please see: http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/features/dedup_disk/tools.htm

Regards,

Phil

On 2012-02-01, at 2:50 PM, Paul Hutchings wrote:
Quote

What is the recommended way to benchmark various RAID configs for maglib volumes please?

Environment will be dedupe with auxcopy to tape.

I know I can use IOMeter, but I'm interested in any specific Commvault tools and what kind of profile (IO size, type, distribution etc.) I should use.

Thanks,
Paul
--
MIRA Ltd

Watling Street, Nuneaton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registration GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

***************************Legal Disclaimer***************************
"This communication may contain confidential and privileged material for the
sole use of the intended recipient. Any unauthorized review, use or distribution
by others is strictly prohibited. If you have received the message in error,
please advise the sender by reply email and delete the message. Thank you."
**********************************************************************

__._,_.___
Reply to sender (pnormand < at > commvault.com?subject=Re%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply to group (commvault < at > yahoogroups.com?subject=Re%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply via web post | Start a New Topic
Messages in this topic (2)
Recent Activity:Visit Your Group

Commvault Documentation here:
http://documentation.commvault.com/
Text-Only (commvault-traditional < at > yahoogroups.com?subject=Change Delivery Format: Traditional), Daily Digest (commvault-digest < at > yahoogroups.com?subject=Email Delivery: Digest) • Unsubscribe (commvault-unsubscribe < at > yahoogroups.com?subject=Unsubscribe) • Terms of Use

.


__,_._,___
Benchmarking Maglib Spindles?
February 01, 2012 12:39PM
Thanks Phil, taking a good look at those. The DDB will be on an SSD. I recall in one of your replies to another of my posts you said along the line that SSD doesn't lend itself to DDB quite the same as you might expect, but from any/all benchmarks that I looked at it should hopefully still murder 3-4 10 or 15k spindles so I'm hopeful.
From: commvault < at > yahoogroups.com [commvault < at > yahoogroups.com] on behalf of Philippe Normand [pnormand < at > commvault.com]
Sent: 01 February 2012 8:10 PM
To: commvault < at > yahoogroups.com
Subject: Re: [commvault] Benchmarking Maglib Spindles?

Hi Paul,

You can use CVDISKPERF and the SIDB2 tools. CVDISKPERF will help you benchmark your maglib while the SIDB2 tool will help you benchmark the disk(s) on which you will put the DDB. By the way, disk performance for the DDB is a lot more critical than the one for the maglib itself. The faster your DDB disk(s) the faster your dedup will run.

For the docs, please see: http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/features/dedup_disk/tools.htm

Regards,

Phil

On 2012-02-01, at 2:50 PM, Paul Hutchings wrote:
Quote

What is the recommended way to benchmark various RAID configs for maglib volumes please?

Environment will be dedupe with auxcopy to tape.

I know I can use IOMeter, but I'm interested in any specific Commvault tools and what kind of profile (IO size, type, distribution etc.) I should use.

Thanks,
Paul
--
MIRA Ltd

Watling Street, Nuneaton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registration GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

***************************Legal Disclaimer***************************
"This communication may contain confidential and privileged material for the
sole use of the intended recipient. Any unauthorized review, use or distribution
by others is strictly prohibited. If you have received the message in error,
please advise the sender by reply email and delete the message. Thank you."
**********************************************************************

__._,_.___
Reply to sender (paul.hutchings &lt; at &gt; mira.co.uk?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply to group (commvault &lt; at &gt; yahoogroups.com?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply via web post | Start a New Topic
Messages in this topic (3)
Recent Activity:Visit Your Group

Commvault Documentation here:
http://documentation.commvault.com/
Text-Only (commvault-traditional &lt; at &gt; yahoogroups.com?subject=Change Delivery Format: Traditional), Daily Digest (commvault-digest &lt; at &gt; yahoogroups.com?subject=Email Delivery: Digest) &bull; Unsubscribe (commvault-unsubscribe &lt; at &gt; yahoogroups.com?subject=Unsubscribe) &bull; Terms of Use

.


__,_._,___
Benchmarking Maglib Spindles?
February 01, 2012 12:49PM
No problem. Let us know how you make out with the SSD. Curious to see how it behaves in the real world.

Regards,

Phil

On 2012-02-01, at 3:39 PM, Paul Hutchings wrote:
Quote

Thanks Phil, taking a good look at those. The DDB will be on an SSD. I recall in one of your replies to another of my posts you said along the line that SSD doesn't lend itself to DDB quite the same as you might expect, but from any/all benchmarks that I looked at it should hopefully still murder 3-4 10 or 15k spindles so I'm hopeful.
From: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com) [commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand &lt; at &gt; commvault.com)]
Sent: 01 February 2012 8:10 PM
To: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

Hi Paul,

You can use CVDISKPERF and the SIDB2 tools. CVDISKPERF will help you benchmark your maglib while the SIDB2 tool will help you benchmark the disk(s) on which you will put the DDB. By the way, disk performance for the DDB is a lot more critical than the one for the maglib itself. The faster your DDB disk(s) the faster your dedup will run.

For the docs, please see: http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/features/dedup_disk/tools.htm

Regards,

Phil

On 2012-02-01, at 2:50 PM, Paul Hutchings wrote:
Quote

What is the recommended way to benchmark various RAID configs for maglib volumes please?

Environment will be dedupe with auxcopy to tape.

I know I can use IOMeter, but I'm interested in any specific Commvault tools and what kind of profile (IO size, type, distribution etc.) I should use.

Thanks,
Paul
--
MIRA Ltd

Watling Street, Nuneaton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registration GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

***************************Legal Disclaimer***************************
"This communication may contain confidential and privileged material for the
sole use of the intended recipient. Any unauthorized review, use or distribution
by others is strictly prohibited. If you have received the message in error,
please advise the sender by reply email and delete the message. Thank you."
**********************************************************************

__._,_.___
Reply to sender (pnormand &lt; at &gt; commvault.com?subject=Re%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply to group (commvault &lt; at &gt; yahoogroups.com?subject=Re%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply via web post | Start a New Topic
Messages in this topic (4)
Recent Activity:Visit Your Group

Commvault Documentation here:
http://documentation.commvault.com/
Text-Only (commvault-traditional &lt; at &gt; yahoogroups.com?subject=Change Delivery Format: Traditional), Daily Digest (commvault-digest &lt; at &gt; yahoogroups.com?subject=Email Delivery: Digest) &bull; Unsubscribe (commvault-unsubscribe &lt; at &gt; yahoogroups.com?subject=Unsubscribe) &bull; Terms of Use

.


__,_._,___
Benchmarking Maglib Spindles?
February 01, 2012 12:52PM
I may be back crying like a baby smiling smiley I'm making some assumptions here but when you guys tested it what were you comparing it against?

We're firmly at the bottom end of the Commvault scale so if we had gone with RAID0 for the DDB we'd only have been able to stretch to 3 or 4.
From: commvault < at > yahoogroups.com [commvault < at > yahoogroups.com] on behalf of Philippe Normand [pnormand < at > commvault.com]
Sent: 01 February 2012 8:49 PM
To: commvault < at > yahoogroups.com
Subject: Re: [commvault] Benchmarking Maglib Spindles?

No problem. Let us know how you make out with the SSD. Curious to see how it behaves in the real world.

Regards,

Phil

On 2012-02-01, at 3:39 PM, Paul Hutchings wrote:
Quote

Thanks Phil, taking a good look at those. The DDB will be on an SSD. I recall in one of your replies to another of my posts you said along the line that SSD doesn't lend itself to DDB quite the same as you might expect, but from any/all benchmarks that I looked at it should hopefully still murder 3-4 10 or 15k spindles so I'm hopeful.
From: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com) [commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand &lt; at &gt; commvault.com)]
Sent: 01 February 2012 8:10 PM
To: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

Hi Paul,

You can use CVDISKPERF and the SIDB2 tools. CVDISKPERF will help you benchmark your maglib while the SIDB2 tool will help you benchmark the disk(s) on which you will put the DDB. By the way, disk performance for the DDB is a lot more critical than the one for the maglib itself. The faster your DDB disk(s) the faster your dedup will run.

For the docs, please see: http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/features/dedup_disk/tools.htm

Regards,

Phil

On 2012-02-01, at 2:50 PM, Paul Hutchings wrote:
Quote

What is the recommended way to benchmark various RAID configs for maglib volumes please?

Environment will be dedupe with auxcopy to tape.

I know I can use IOMeter, but I'm interested in any specific Commvault tools and what kind of profile (IO size, type, distribution etc.) I should use.

Thanks,
Paul
--
MIRA Ltd

Watling Street, Nuneaton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registration GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

***************************Legal Disclaimer***************************
"This communication may contain confidential and privileged material for the
sole use of the intended recipient. Any unauthorized review, use or distribution
by others is strictly prohibited. If you have received the message in error,
please advise the sender by reply email and delete the message. Thank you."
**********************************************************************

__._,_.___
Reply to sender (paul.hutchings &lt; at &gt; mira.co.uk?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply to group (commvault &lt; at &gt; yahoogroups.com?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply via web post | Start a New Topic
Messages in this topic (5)
Recent Activity:Visit Your Group

Commvault Documentation here:
http://documentation.commvault.com/
Text-Only (commvault-traditional &lt; at &gt; yahoogroups.com?subject=Change Delivery Format: Traditional), Daily Digest (commvault-digest &lt; at &gt; yahoogroups.com?subject=Email Delivery: Digest) &bull; Unsubscribe (commvault-unsubscribe &lt; at &gt; yahoogroups.com?subject=Unsubscribe) &bull; Terms of Use

.


__,_._,___
Benchmarking Maglib Spindles?
February 01, 2012 12:56PM
What SSD are you using ? On Feb 1, 2012 3:52 PM, "Paul Hutchings" <paul.hutchings < at > mira.co.uk (paul.hutchings &lt; at &gt; mira.co.uk)> wrote:
Quote

 
I may be back crying like a baby smiling smiley  I&#39;m making some assumptions here but when you guys tested it what were you comparing it against?

We&#39;re firmly at the bottom end of the Commvault scale so if we had gone with RAID0 for the DDB we&#39;d only have been able to stretch to 3 or 4.
From: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com) [commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand &lt; at &gt; commvault.com)]
Sent: 01 February 2012 8:49 PM
To: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

 
No problem. Let us know how you make out with the SSD. Curious to see how it behaves in the real world.

Regards, 

Phil

On 2012-02-01, at 3:39 PM, Paul Hutchings wrote:

Quote

 

Thanks Phil, taking a good look at those.  The DDB will be on an SSD.  I recall in one of your replies to another of my posts you said along the line that SSD doesn&#39;t lend itself to DDB quite the same as you might expect, but from any/all benchmarks that I looked at it should hopefully still murder 3-4 10 or 15k spindles so I&#39;m hopeful.
From: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com) [commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand &lt; at &gt; commvault.com)]
Sent: 01 February 2012 8:10 PM
To: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

 
Hi Paul, 

You can use CVDISKPERF and the SIDB2 tools. CVDISKPERF will help you benchmark your maglib while the SIDB2 tool will help you benchmark the disk(s) on which you will put the DDB. By the way, disk performance for the DDB is a lot more critical than the one for the maglib itself. The faster your DDB disk(s) the faster your dedup will run. 

For the docs, please see: http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/features/dedup_disk/tools.htm

Regards, 

Phil

On 2012-02-01, at 2:50 PM, Paul Hutchings wrote:

Quote

 
What is the recommended way to benchmark various RAID configs for maglib volumes please?

Environment will be dedupe with auxcopy to tape.

I know I can use IOMeter, but I&#39;m interested in any specific Commvault tools and what kind of profile (IO size, type, distribution etc.) I should use.

Thanks,
Paul
--
MIRA Ltd

Watling Street, Nuneaton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registration GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

***************************Legal Disclaimer***************************
"This communication may contain confidential and privileged material for the
sole use of the intended recipient. Any unauthorized review, use or distribution
by others is strictly prohibited. If you have received the message in error,
please advise the sender by reply email and delete the message. Thank you."
**********************************************************************

__._,_.___
Reply to sender (kelly.brianx &lt; at &gt; gmail.com?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply to group (commvault &lt; at &gt; yahoogroups.com?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply via web post | Start a New Topic
Messages in this topic (6)
Recent Activity:Visit Your Group

Commvault Documentation here:
http://documentation.commvault.com/
Text-Only (commvault-traditional &lt; at &gt; yahoogroups.com?subject=Change Delivery Format: Traditional), Daily Digest (commvault-digest &lt; at &gt; yahoogroups.com?subject=Email Delivery: Digest) &bull; Unsubscribe (commvault-unsubscribe &lt; at &gt; yahoogroups.com?subject=Unsubscribe) &bull; Terms of Use

.


__,_._,___
Benchmarking Maglib Spindles?
February 01, 2012 12:58PM
Intel 320. From some feedback from some other folks who were using them they seemed solid/reliable enough.
From: commvault < at > yahoogroups.com [commvault < at > yahoogroups.com] on behalf of Brian Kelly [kelly.brianx < at > gmail.com]
Sent: 01 February 2012 8:56 PM
To: commvault < at > yahoogroups.com
Subject: RE: [commvault] Benchmarking Maglib Spindles?

What SSD are you using ? On Feb 1, 2012 3:52 PM, "Paul Hutchings" <paul.hutchings < at > mira.co.uk (paul.hutchings &lt; at &gt; mira.co.uk)> wrote:
Quote


I may be back crying like a baby smiling smiley I'm making some assumptions here but when you guys tested it what were you comparing it against?

We're firmly at the bottom end of the Commvault scale so if we had gone with RAID0 for the DDB we'd only have been able to stretch to 3 or 4.
From: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com) [commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand &lt; at &gt; commvault.com)]
Sent: 01 February 2012 8:49 PM
To: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

No problem. Let us know how you make out with the SSD. Curious to see how it behaves in the real world.

Regards,

Phil

On 2012-02-01, at 3:39 PM, Paul Hutchings wrote:

Quote


Thanks Phil, taking a good look at those. The DDB will be on an SSD. I recall in one of your replies to another of my posts you said along the line that SSD doesn't lend itself to DDB quite the same as you might expect, but from any/all benchmarks that I looked at it should hopefully still murder 3-4 10 or 15k spindles so I'm hopeful.
From: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com) [commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand &lt; at &gt; commvault.com)]
Sent: 01 February 2012 8:10 PM
To: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

Hi Paul,

You can use CVDISKPERF and the SIDB2 tools. CVDISKPERF will help you benchmark your maglib while the SIDB2 tool will help you benchmark the disk(s) on which you will put the DDB. By the way, disk performance for the DDB is a lot more critical than the one for the maglib itself. The faster your DDB disk(s) the faster your dedup will run.

For the docs, please see: http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/features/dedup_disk/tools.htm

Regards,

Phil

On 2012-02-01, at 2:50 PM, Paul Hutchings wrote:

Quote

What is the recommended way to benchmark various RAID configs for maglib volumes please?

Environment will be dedupe with auxcopy to tape.

I know I can use IOMeter, but I'm interested in any specific Commvault tools and what kind of profile (IO size, type, distribution etc.) I should use.

Thanks,
Paul
--
MIRA Ltd

Watling Street, Nuneaton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registration GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

***************************Legal Disclaimer***************************
"This communication may contain confidential and privileged material for the
sole use of the intended recipient. Any unauthorized review, use or distribution
by others is strictly prohibited. If you have received the message in error,
please advise the sender by reply email and delete the message. Thank you."
**********************************************************************

__._,_.___
Reply to sender (paul.hutchings &lt; at &gt; mira.co.uk?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply to group (commvault &lt; at &gt; yahoogroups.com?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply via web post | Start a New Topic
Messages in this topic (7)
Recent Activity:Visit Your Group

Commvault Documentation here:
http://documentation.commvault.com/
Text-Only (commvault-traditional &lt; at &gt; yahoogroups.com?subject=Change Delivery Format: Traditional), Daily Digest (commvault-digest &lt; at &gt; yahoogroups.com?subject=Email Delivery: Digest) &bull; Unsubscribe (commvault-unsubscribe &lt; at &gt; yahoogroups.com?subject=Unsubscribe) &bull; Terms of Use

.


__,_._,___
Benchmarking Maglib Spindles?
February 01, 2012 01:12PM
I am a big fan of SSDs. Spinning disk performance has not kept up with the rest of the technologies like CPUs and memory. 10/15K drives are not going to survive in the long run...

Unfortunately, no detailed info was provided when we discussed the benchmarks. However, since the SIDB process can heavily load the device with a lot of random writes, and as much as TRIM does miracles to hide SSD write limitations (read/compare/write/verify), after a while you may run into issues like Write Amplification (http://en.wikipedia.org/wiki/Write_amplification) which ends up killing the performance down to a good old spinning disk... That's probably the point where you start crying ;-)

On 2012-02-01, at 3:56 PM, Brian Kelly wrote:
Quote


What SSD are you using ? On Feb 1, 2012 3:52 PM, "Paul Hutchings" <paul.hutchings < at > mira.co.uk (paul.hutchings &lt; at &gt; mira.co.uk)> wrote:
Quote


I may be back crying like a baby smiling smiley I'm making some assumptions here but when you guys tested it what were you comparing it against?

We're firmly at the bottom end of the Commvault scale so if we had gone with RAID0 for the DDB we'd only have been able to stretch to 3 or 4.
From: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com) [commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand &lt; at &gt; commvault.com)]
Sent: 01 February 2012 8:49 PM
To: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

No problem. Let us know how you make out with the SSD. Curious to see how it behaves in the real world.

Regards,

Phil

On 2012-02-01, at 3:39 PM, Paul Hutchings wrote:

Quote


Thanks Phil, taking a good look at those. The DDB will be on an SSD. I recall in one of your replies to another of my posts you said along the line that SSD doesn't lend itself to DDB quite the same as you might expect, but from any/all benchmarks that I looked at it should hopefully still murder 3-4 10 or 15k spindles so I'm hopeful.
From: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com) [commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand &lt; at &gt; commvault.com)]
Sent: 01 February 2012 8:10 PM
To: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

Hi Paul,

You can use CVDISKPERF and the SIDB2 tools. CVDISKPERF will help you benchmark your maglib while the SIDB2 tool will help you benchmark the disk(s) on which you will put the DDB. By the way, disk performance for the DDB is a lot more critical than the one for the maglib itself. The faster your DDB disk(s) the faster your dedup will run.

For the docs, please see: http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/features/dedup_disk/tools.htm

Regards,

Phil

On 2012-02-01, at 2:50 PM, Paul Hutchings wrote:

Quote

What is the recommended way to benchmark various RAID configs for maglib volumes please?

Environment will be dedupe with auxcopy to tape.

I know I can use IOMeter, but I'm interested in any specific Commvault tools and what kind of profile (IO size, type, distribution etc.) I should use.

Thanks,
Paul
--
MIRA Ltd

Watling Street, Nuneaton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registration GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

***************************Legal Disclaimer***************************
"This communication may contain confidential and privileged material for the
sole use of the intended recipient. Any unauthorized review, use or distribution
by others is strictly prohibited. If you have received the message in error,
please advise the sender by reply email and delete the message. Thank you."
**********************************************************************

__._,_.___
Reply to sender (pnormand &lt; at &gt; commvault.com?subject=Re%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply to group (commvault &lt; at &gt; yahoogroups.com?subject=Re%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply via web post | Start a New Topic
Messages in this topic (8)
Recent Activity:Visit Your Group

Commvault Documentation here:
http://documentation.commvault.com/
Text-Only (commvault-traditional &lt; at &gt; yahoogroups.com?subject=Change Delivery Format: Traditional), Daily Digest (commvault-digest &lt; at &gt; yahoogroups.com?subject=Email Delivery: Digest) &bull; Unsubscribe (commvault-unsubscribe &lt; at &gt; yahoogroups.com?subject=Unsubscribe) &bull; Terms of Use

.


__,_._,___
Benchmarking Maglib Spindles?
February 01, 2012 01:23PM
See now I could lie and say I chose the Intel 320 because the documentations explicitly claims low write amplification :-)

Truth is though, I chose it because it seemed to get decent reviews as a solid performer, and a couple of folks reported first hand using them an with a Dell PERC RAID controller without any issues.

When you say "after a while" with regard to WA, do you mean as in if I ran IOMeter you'd expect to see performance drop after minutes of sustained usage, or are you talking specifically long term? That's where the Wikipedia article started to get a little heavy.
From: commvault < at > yahoogroups.com [commvault < at > yahoogroups.com] on behalf of Philippe Normand [pnormand < at > commvault.com]
Sent: 01 February 2012 9:12 PM
To: commvault < at > yahoogroups.com
Subject: Re: [commvault] Benchmarking Maglib Spindles?

I am a big fan of SSDs. Spinning disk performance has not kept up with the rest of the technologies like CPUs and memory. 10/15K drives are not going to survive in the long run...

Unfortunately, no detailed info was provided when we discussed the benchmarks. However, since the SIDB process can heavily load the device with a lot of random writes, and as much as TRIM does miracles to hide SSD write limitations (read/compare/write/verify), after a while you may run into issues like Write Amplification (http://en.wikipedia.org/wiki/Write_amplification) which ends up killing the performance down to a good old spinning disk... That's probably the point where you start crying ;-)

On 2012-02-01, at 3:56 PM, Brian Kelly wrote:
Quote


What SSD are you using ? On Feb 1, 2012 3:52 PM, "Paul Hutchings" <paul.hutchings < at > mira.co.uk (paul.hutchings &lt; at &gt; mira.co.uk)> wrote:
Quote


I may be back crying like a baby smiling smiley I'm making some assumptions here but when you guys tested it what were you comparing it against?

We're firmly at the bottom end of the Commvault scale so if we had gone with RAID0 for the DDB we'd only have been able to stretch to 3 or 4.
From: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com) [commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand &lt; at &gt; commvault.com)]
Sent: 01 February 2012 8:49 PM
To: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

No problem. Let us know how you make out with the SSD. Curious to see how it behaves in the real world.

Regards,

Phil

On 2012-02-01, at 3:39 PM, Paul Hutchings wrote:

Quote


Thanks Phil, taking a good look at those. The DDB will be on an SSD. I recall in one of your replies to another of my posts you said along the line that SSD doesn't lend itself to DDB quite the same as you might expect, but from any/all benchmarks that I looked at it should hopefully still murder 3-4 10 or 15k spindles so I'm hopeful.
From: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com) [commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand &lt; at &gt; commvault.com)]
Sent: 01 February 2012 8:10 PM
To: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

Hi Paul,

You can use CVDISKPERF and the SIDB2 tools. CVDISKPERF will help you benchmark your maglib while the SIDB2 tool will help you benchmark the disk(s) on which you will put the DDB. By the way, disk performance for the DDB is a lot more critical than the one for the maglib itself. The faster your DDB disk(s) the faster your dedup will run.

For the docs, please see: http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/features/dedup_disk/tools.htm

Regards,

Phil

On 2012-02-01, at 2:50 PM, Paul Hutchings wrote:

Quote

What is the recommended way to benchmark various RAID configs for maglib volumes please?

Environment will be dedupe with auxcopy to tape.

I know I can use IOMeter, but I'm interested in any specific Commvault tools and what kind of profile (IO size, type, distribution etc.) I should use.

Thanks,
Paul
--
MIRA Ltd

Watling Street, Nuneaton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registration GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

***************************Legal Disclaimer***************************
"This communication may contain confidential and privileged material for the
sole use of the intended recipient. Any unauthorized review, use or distribution
by others is strictly prohibited. If you have received the message in error,
please advise the sender by reply email and delete the message. Thank you."
**********************************************************************

__._,_.___
Reply to sender (paul.hutchings &lt; at &gt; mira.co.uk?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply to group (commvault &lt; at &gt; yahoogroups.com?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply via web post | Start a New Topic
Messages in this topic (9)
Recent Activity:Visit Your Group

Commvault Documentation here:
http://documentation.commvault.com/
Text-Only (commvault-traditional &lt; at &gt; yahoogroups.com?subject=Change Delivery Format: Traditional), Daily Digest (commvault-digest &lt; at &gt; yahoogroups.com?subject=Email Delivery: Digest) &bull; Unsubscribe (commvault-unsubscribe &lt; at &gt; yahoogroups.com?subject=Unsubscribe) &bull; Terms of Use

.


__,_._,___
Benchmarking Maglib Spindles?
February 14, 2012 08:36AM
For what it's worth, SSD is here, just used cvdiskperf to do a quick benchmark:

DiskPerf Version : 1.3
Path Used : Econfused smileyIDB01
Read-Write type : RANDOM
Block Size : 512
Block Count : 4096
File Count : 1024
Total Bytes Written : 2147483648
Time Taken to Write(S) : 9.89
Throughput Write(GB/H) : 727.82
Total Bytes Read : 2147483648
Time Taken to Read(S) : 8.56
Throughput Read(GB/H) : 841.13
Time Taken to Create(S) : 7.90
Throughput Create(GB/H) : 911.15

I haven't yet installed Commvault directly on the MA (I just copied the base folder) but if there's any more DDB specific benchmarks I'd be interested - the ones in BOL seemed to need the instancexxx to run?

Paul

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Paul Hutchings
Sent: 01 February 2012 21:24
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?

 
See now I could lie and say I chose the Intel 320 because the documentations explicitly claims low write amplification :-)

Truth is though, I chose it because it seemed to get decent reviews as a solid performer, and a couple of folks reported first hand using them an with a Dell PERC RAID controller without any issues.

When you say "after a while" with regard to WA, do you mean as in if I ran IOMeter you'd expect to see performance drop after minutes of sustained usage, or are you talking specifically long term?  That's where the Wikipedia article started to get a little heavy.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 9:12 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?
 
I am a big fan of SSDs. Spinning disk performance has not kept up with the rest of the technologies like CPUs and memory. 10/15K drives are not going to survive in the long run...

Unfortunately, no detailed info was provided when we discussed the benchmarks. However, since the SIDB process can heavily load the device with a lot of random writes, and as much as TRIM does miracles to hide SSD write limitations (read/compare/write/verify), after a while you may run into issues like Write Amplification (http://en.wikipedia.org/wiki/Write_amplification) which ends up killing the performance down to a good old spinning disk... That's probably the point where you start crying ;-)

On 2012-02-01, at 3:56 PM, Brian Kelly wrote:

 

What SSD are you using ?
On Feb 1, 2012 3:52 PM, "Paul Hutchings" <paul.hutchings < at > mira.co.uk (paul.hutchings%40mira.co.uk)> wrote:
 

I may be back crying like a baby smiling smiley  I'm making some assumptions here but when you guys tested it what were you comparing it against?

We're firmly at the bottom end of the Commvault scale so if we had gone with RAID0 for the DDB we'd only have been able to stretch to 3 or 4.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 8:49 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?
 
No problem. Let us know how you make out with the SSD. Curious to see how it behaves in the real world.

Regards, 

Phil

On 2012-02-01, at 3:39 PM, Paul Hutchings wrote:

 

Thanks Phil, taking a good look at those.  The DDB will be on an SSD.  I recall in one of your replies to another of my posts you said along the line that SSD doesn't lend itself to DDB quite the same as you might expect, but from any/all benchmarks that I looked at it should hopefully still murder 3-4 10 or 15k spindles so I'm hopeful.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 8:10 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?
 
Hi Paul, 

You can use CVDISKPERF and the SIDB2 tools. CVDISKPERF will help you benchmark your maglib while the SIDB2 tool will help you benchmark the disk(s) on which you will put the DDB. By the way, disk performance for the DDB is a lot more critical than the one for the maglib itself. The faster your DDB disk(s) the faster your dedup will run. 

For the docs, please see: http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/features/dedup_disk/tools.htm

Regards, 

Phil

On 2012-02-01, at 2:50 PM, Paul Hutchings wrote:

 
What is the recommended way to benchmark various RAID configs for maglib volumes please?

Environment will be dedupe with auxcopy to tape.

I know I can use IOMeter, but I'm interested in any specific Commvault tools and what kind of profile (IO size, type, distribution etc.) I should use.

Thanks,
Paul
--
MIRA Ltd

Watling Street, Nuneaton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registration GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

***************************Legal Disclaimer***************************
"This communication may contain confidential and privileged material for the
sole use of the intended recipient. Any unauthorized review, use or distribution
by others is strictly prohibited. If you have received the message in error,
please advise the sender by reply email and delete the message. Thank you."
**********************************************************************

__._,_.___
Reply to sender (paul.hutchings &lt; at &gt; mira.co.uk?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply to group (commvault &lt; at &gt; yahoogroups.com?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply via web post | Start a New Topic
Messages in this topic (10)
Recent Activity:Visit Your Group

Commvault Documentation here:
http://documentation.commvault.com/
Text-Only (commvault-traditional &lt; at &gt; yahoogroups.com?subject=Change Delivery Format: Traditional), Daily Digest (commvault-digest &lt; at &gt; yahoogroups.com?subject=Email Delivery: Digest) &bull; Unsubscribe (commvault-unsubscribe &lt; at &gt; yahoogroups.com?subject=Unsubscribe) &bull; Terms of Use

.


__,_._,___
Benchmarking Maglib Spindles?
February 14, 2012 10:12PM
Hey Paul,

You should be running a simulateddb from SIDB2 to get a more specific DDB benchmark:

http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/prod_info/dedup_disk.htm?var1=http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/features/dedup_disk/tools.htm
Sample command:

SIDB2.exe simulateddb p Econfused smileyimulate in Instance001 datasize 2048

Would be interested to see what your results are against the SSD J What model/manufacturer of SSD are you using.. Intel?

Cheers,
Luke

From: commvault < at > yahoogroups.com [mailto] On Behalf Of Paul Hutchings
Sent: Wednesday, 15 February 2012 3:36 AM
To: commvault < at > yahoogroups.com
Subject: RE: [commvault] Benchmarking Maglib Spindles?

For what it's worth, SSD is here, just used cvdiskperf to do a quick benchmark:

DiskPerf Version : 1.3
Path Used : Econfused smileyIDB01
Read-Write type : RANDOM
Block Size : 512
Block Count : 4096
File Count : 1024
Total Bytes Written : 2147483648
Time Taken to Write(S) : 9.89
Throughput Write(GB/H) : 727.82
Total Bytes Read : 2147483648
Time Taken to Read(S) : 8.56
Throughput Read(GB/H) : 841.13
Time Taken to Create(S) : 7.90
Throughput Create(GB/H) : 911.15

I haven't yet installed Commvault directly on the MA (I just copied the base folder) but if there's any more DDB specific benchmarks I'd be interested - the ones in BOL seemed to need the instancexxx to run?

Paul

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Paul Hutchings
Sent: 01 February 2012 21:24
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?

See now I could lie and say I chose the Intel 320 because the documentations explicitly claims low write amplification :-)

Truth is though, I chose it because it seemed to get decent reviews as a solid performer, and a couple of folks reported first hand using them an with a Dell PERC RAID controller without any issues.

When you say "after a while" with regard to WA, do you mean as in if I ran IOMeter you'd expect to see performance drop after minutes of sustained usage, or are you talking specifically long term? That's where the Wikipedia article started to get a little heavy.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 9:12 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

I am a big fan of SSDs. Spinning disk performance has not kept up with the rest of the technologies like CPUs and memory. 10/15K drives are not going to survive in the long run...

Unfortunately, no detailed info was provided when we discussed the benchmarks. However, since the SIDB process can heavily load the device with a lot of random writes, and as much as TRIM does miracles to hide SSD write limitations (read/compare/write/verify), after a while you may run into issues like Write Amplification (http://en.wikipedia.org/wiki/Write_amplification) which ends up killing the performance down to a good old spinning disk... That's probably the point where you start crying ;-)

On 2012-02-01, at 3:56 PM, Brian Kelly wrote:

What SSD are you using ?
On Feb 1, 2012 3:52 PM, "Paul Hutchings" <paul.hutchings < at > mira.co.uk (paul.hutchings%40mira.co.uk)> wrote:

I may be back crying like a baby smiling smiley I'm making some assumptions here but when you guys tested it what were you comparing it against?

We're firmly at the bottom end of the Commvault scale so if we had gone with RAID0 for the DDB we'd only have been able to stretch to 3 or 4.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 8:49 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

No problem. Let us know how you make out with the SSD. Curious to see how it behaves in the real world.

Regards,

Phil

On 2012-02-01, at 3:39 PM, Paul Hutchings wrote:

Thanks Phil, taking a good look at those. The DDB will be on an SSD. I recall in one of your replies to another of my posts you said along the line that SSD doesn't lend itself to DDB quite the same as you might expect, but from any/all benchmarks that I looked at it should hopefully still murder 3-4 10 or 15k spindles so I'm hopeful.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 8:10 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

Hi Paul,

You can use CVDISKPERF and the SIDB2 tools. CVDISKPERF will help you benchmark your maglib while the SIDB2 tool will help you benchmark the disk(s) on which you will put the DDB. By the way, disk performance for the DDB is a lot more critical than the one for the maglib itself. The faster your DDB disk(s) the faster your dedup will run.

For the docs, please see: http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/features/dedup_disk/tools.htm

Regards,

Phil

On 2012-02-01, at 2:50 PM, Paul Hutchings wrote:

What is the recommended way to benchmark various RAID configs for maglib volumes please?

Environment will be dedupe with auxcopy to tape.

I know I can use IOMeter, but I'm interested in any specific Commvault tools and what kind of profile (IO size, type, distribution etc.) I should use.

Thanks,
Paul
--
MIRA Ltd

Watling Street, Nuneaton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registration GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

***************************Legal Disclaimer***************************
"This communication may contain confidential and privileged material for the
sole use of the intended recipient. Any unauthorized review, use or distribution
by others is strictly prohibited. If you have received the message in error,
please advise the sender by reply email and delete the message. Thank you."
**********************************************************************

__._,_.___
Reply to sender (luke &lt; at &gt; blackduck.nu?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply to group (commvault &lt; at &gt; yahoogroups.com?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply via web post | Start a New Topic
Messages in this topic (11)
Recent Activity:Visit Your Group

Commvault Documentation here:
http://documentation.commvault.com/
Text-Only (commvault-traditional &lt; at &gt; yahoogroups.com?subject=Change Delivery Format: Traditional), Daily Digest (commvault-digest &lt; at &gt; yahoogroups.com?subject=Email Delivery: Digest) &bull; Unsubscribe (commvault-unsubscribe &lt; at &gt; yahoogroups.com?subject=Unsubscribe) &bull; Terms of Use

.


__,_._,___
Benchmarking Maglib Spindles?
February 15, 2012 02:17AM
Thanks Luke, that&#8217;s been running for an hour and a half or so using SIDB2 -simulateddb -in instance001 -p e:dedup_store.

The SSD is an Intel 320. I&#8217;ll report back once I have some numbers.

From: commvault < at > yahoogroups.com [mailto] On Behalf Of Luke Walker
Sent: 14 February 2012 20:23
To: commvault < at > yahoogroups.com
Subject: RE: [commvault] Benchmarking Maglib Spindles?

Hey Paul,

You should be running a simulateddb from SIDB2 to get a more specific DDB benchmark:

http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/prod_info/dedup_disk.htm?var1=http://documentation.commvault.com/commvault/release_9_0_0/books_online_1! /english_us/features/dedup_disk/tools.htm
Sample command:

SIDB2.exe simulateddb p Econfused smileyimulate in Instance001 datasize 2

Would be interested to see what your results are against the SSD J What model/manufacturer of SSD are you using.. Intel?

Cheers,
Luke

From: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com) [mailto] ([mailto]) On Behalf Of Paul Hutchings
Sent: Wednesday, 15 February 2012 3:36 AM
To: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?

For what it's worth, SSD is here, just used cvdiskperf to do a quick benchmark:

DiskPerf Version : 1.3
Path Used : Econfused smileyIDB01
Read-Write type : RANDOM
Block ! Size : 512
Block Count : 4096
File Count : 1024
Total Bytes Written : 2147483648
Time Taken to Write(S) : 9.89
Throughput Write(GB/H) : 727.82
Total Bytes Read : 2147483648
Time Taken to Read(S) : 8.56
Throughput Read(GB/H) : 841.13
Time Taken to Create(S) : 7.90
Throughput Create(GB/H) : 911.15

I haven't yet installed Commvault directly on the MA (I just copied the base folder) but if there's any more DDB specific benchmarks I'd be interested - the ones in BOL seemed to need the instancexxx to run?

Paul

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Paul Hutchings
Sent: 01 February 2012 21:24
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?

See now I could lie and say I chose the Inte! l 320 because the documentations explicitly claims low write amplifica tion :-)

Truth is though, I chose it because it seemed to get decent reviews as a solid performer, and a couple of folks reported first hand using them an with a Dell PERC RAID controller without any issues.

When you say "after a while" with regard to WA, do you mean as in if I ran IOMeter you'd expect to see performance drop after minutes of sustained usage, or are you talking specifically long term? That's where the Wikipedia article started to get a little heavy.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 9:12 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?
! ;
I am a big fan of SSDs. Spinning disk performance has not kept up with the rest of the technologies like CPUs and memory. 10/15K drives are not going to survive in the long run...

Unfortunately, no detailed info was provided when we discussed the benchmarks. However, since the SIDB process can heavily load the device with a lot of random writes, and as much as TRIM does miracles to hide SSD write limitations (read/compare/write/verify), after a while you may run into issues like Write Amplification (http://en.wikipedia.org/wiki/Write_amplification) which ends up killing the performance down to a good old spinning disk... That's probably the point where you start crying ;-)

On 2012-02-01, at 3:56 PM, Brian Kelly wrote:

What SSD are you using ?
On Feb 1, 2012 3:52 PM, "Paul Hutchings" <paul.hutchings < at > mira.c! o.uk (paul.hutchings%40mira.co.uk)> wrote:

I may be back crying like a baby smiling smiley I'm making some assumptions here but when you guys tested it what were you comparing it against?

We're firmly at the bottom end of the Commvault scale so if we had gone with RAID0 for the DDB we'd only have been able to stretch to 3 or 4.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 8:49 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

No problem. Let us know how you make out with the SSD. Curious to see how it behaves in the real world.

Regards,

Phil

On 2012-02-01, at 3:39 PM, Paul Hutchings wrote:

Thanks Phil, taking a ! good look at those. The DDB will be on an SSD. I recall in one of your replies to another of my posts you said along the line that SSD doesn't lend itself to DDB quite the same as you might expect, but from any/all benchmarks that I looked at it should hopefully still murder 3-4 10 or 15k spindles so I'm hopeful.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 8:10 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

Hi Paul,

You can use CVDISKPERF and the SIDB2 tools. CVDISKPERF will help you benchmark your maglib while the SIDB2 tool will help you benchmark the disk(! s) on which you will put the DDB. By the way, disk performance for the DDB is a lot more critical than the one for the maglib itself. The faster your DDB disk(s) the faster your dedup will run.

For the docs, please see: http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/features/dedup_disk/tools.htm

Regards,

Phil

On 2012-02-01, at 2:50 PM, Paul Hutchings wrote:

What is the recommended way to benchmark various RAID configs for maglib volumes please?

Environment will be dedupe with auxcopy to tape.

I know I can use IOMeter, but I'm interested in any specific Commvault tools and what kind of profile (IO size, type, distribution etc.) I should use.

Thanks,
Paul
--
MIRA Ltd

Watling Street, Nuneaton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registr! ation GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

***************************Legal Disclaimer***************************
"This communication may contain confidential and privileged material for the
sole use of the intended recipient. Any unauthorized review, use or distribution
by others is strictly prohibited. If you have received the message in error,
please advise the sender by reply email and delete the message. Thank you."
**********************************************************************

MIRA Ltd

Watling Street, Nuneaton, Warwickshire, CV10 0TU, England

Registered in England and Wales No. 402570
VAT Registration  GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient.  If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax.  You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

__._,_.___
Reply to sender (paul.hutchings &lt; at &gt; mira.co.uk?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply to group (commvault &lt; at &gt; yahoogroups.com?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply via web post | Start a New Topic
Messages in this topic (12)
Recent Activity:Visit Your Group

Commvault Documentation here:
http://documentation.commvault.com/
Text-Only (commvault-traditional &lt; at &gt; yahoogroups.com?subject=Change Delivery Format: Traditional), Daily Digest (commvault-digest &lt; at &gt; yahoogroups.com?subject=Email Delivery: Digest) &bull; Unsubscribe (commvault-unsubscribe &lt; at &gt; yahoogroups.com?subject=Unsubscribe) &bull; Terms of Use

.


__,_._,___
Benchmarking Maglib Spindles?
February 15, 2012 05:11AM
Thanks, looking forward to the numbers.

Be warned it can take a while with no additional options as it&#8217;ll run as hard as it can until it hits its thresholds.. J

You can use the datasize switch (i.e. datasize 500, to simulate 0.5 TB of Application Data, or set it to something larger) to cut down the length of time it takes to run the test.

Cheers,
Luke

From: commvault < at > yahoogroups.com [mailto] On Behalf Of Paul Hutchings
Sent: Wednesday, 15 February 2012 9:17 PM
To: 'commvault < at > yahoogroups.com'
Subject: RE: [commvault] Benchmarking Maglib Spindles?

Thanks Luke, that&#8217;s been running for an hour and a half or so using SIDB2 -simulateddb -in instance001 -p e:dedup_store.

The SSD is an Intel 320. I&#8217;ll report back once I have some numbers.

From: commvault < at > yahoogroups.com [mailto] On Behalf Of Luke Walker
Sent: 14 February 2012 20:23
To: commvault < at > yahoogroups.com
Subject: RE: [commvault] Benchmarking Maglib Spindles?

Hey Paul,

You should be running a simulateddb from SIDB2 to get a more specific DDB benchmark:

http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/prod_info/dedup_disk.htm?var1=http://documentation.commvault.com/commvault/release_9_0_0/books_online_1! /english_us/features/dedup_disk/tools.htm
Sample command:

SIDB2.exe simulateddb p Econfused smileyimulate in Instance001 datasize 2

Would be interested to see what your results are against the SSD J What model/manufacturer of SSD are you using.. Intel?

Cheers,
Luke

From: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com) [mailto] ([mailto]) On Behalf Of Paul Hutchings
Sent: Wednesday, 15 February 2012 3:36 AM
To: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?

For what it's worth, SSD is here, just used cvdiskperf to do a quick benchmark:

DiskPerf Version : 1.3
Path Used : Econfused smileyIDB01
Read-Write type : RANDOM
Block ! Size : 512
Block Count : 4096
File Count : 1024
Total Bytes Written : 2147483648
Time Taken to Write(S) : 9.89
Throughput Write(GB/H) : 727.82
Total Bytes Read : 2147483648
Time Taken to Read(S) : 8.56
Throughput Read(GB/H) : 841.13
Time Taken to Create(S) : 7.90
Throughput Create(GB/H) : 911.15

I haven't yet installed Commvault directly on the MA (I just copied the base folder) but if there's any more DDB specific benchmarks I'd be interested - the ones in BOL seemed to need the instancexxx to run?

Paul

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Paul Hutchings
Sent: 01 February 2012 21:24
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?

See now I could lie and say I chose the Inte! l 320 because the documentations explicitly claims low write amplifica tion :-)

Truth is though, I chose it because it seemed to get decent reviews as a solid performer, and a couple of folks reported first hand using them an with a Dell PERC RAID controller without any issues.

When you say "after a while" with regard to WA, do you mean as in if I ran IOMeter you'd expect to see performance drop after minutes of sustained usage, or are you talking specifically long term? That's where the Wikipedia article started to get a little heavy.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 9:12 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?
! ;
I am a big fan of SSDs. Spinning disk performance has not kept up with the rest of the technologies like CPUs and memory. 10/15K drives are not going to survive in the long run...

Unfortunately, no detailed info was provided when we discussed the benchmarks. However, since the SIDB process can heavily load the device with a lot of random writes, and as much as TRIM does miracles to hide SSD write limitations (read/compare/write/verify), after a while you may run into issues like Write Amplification (http://en.wikipedia.org/wiki/Write_amplification) which ends up killing the performance down to a good old spinning disk... That's probably the point where you start crying ;-)

On 2012-02-01, at 3:56 PM, Brian Kelly wrote:

What SSD are you using ?
On Feb 1, 2012 3:52 PM, "Paul Hutchings" <paul.hutchings < at > mira.c! o.uk (paul.hutchings%40mira.co.uk)> wrote:

I may be back crying like a baby smiling smiley I'm making some assumptions here but when you guys tested it what were you comparing it against?

We're firmly at the bottom end of the Commvault scale so if we had gone with RAID0 for the DDB we'd only have been able to stretch to 3 or 4.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 8:49 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

No problem. Let us know how you make out with the SSD. Curious to see how it behaves in the real world.

Regards,

Phil

On 2012-02-01, at 3:39 PM, Paul Hutchings wrote:

Thanks Phil, taking a ! good look at those. The DDB will be on an SSD. I recall in one of your replies to another of my posts you said along the line that SSD doesn't lend itself to DDB quite the same as you might expect, but from any/all benchmarks that I looked at it should hopefully still murder 3-4 10 or 15k spindles so I'm hopeful.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 8:10 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

Hi Paul,

You can use CVDISKPERF and the SIDB2 tools. CVDISKPERF will help you benchmark your maglib while the SIDB2 tool will help you benchmark the disk(! s) on which you will put the DDB. By the way, disk performance for the DDB is a lot more critical than the one for the maglib itself. The faster your DDB disk(s) the faster your dedup will run.

For the docs, please see: http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/features/dedup_disk/tools.htm

Regards,

Phil

On 2012-02-01, at 2:50 PM, Paul Hutchings wrote:

What is the recommended way to benchmark various RAID configs for maglib volumes please?

Environment will be dedupe with auxcopy to tape.

I know I can use IOMeter, but I'm interested in any specific Commvault tools and what kind of profile (IO size, type, distribution etc.) I should use.

Thanks,
Paul
--
MIRA Ltd

Watling Street, Nuneaton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registr! ation GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

***************************Legal Disclaimer***************************
"This communication may contain confidential and privileged material for the
sole use of the intended recipient. Any unauthorized review, use or distribution
by others is strictly prohibited. If you have received the message in error,
please advise the sender by reply email and delete the message. Thank you."
**********************************************************************

MIRA Ltd

Watling Street, Nuneaton, Warwickshire, CV10 0TU, England

Registered in England and Wales No. 402570

VAT Registration GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

__._,_.___
Reply to sender (luke &lt; at &gt; blackduck.nu?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply to group (commvault &lt; at &gt; yahoogroups.com?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply via web post | Start a New Topic
Messages in this topic (13)
Recent Activity:Visit Your Group

Commvault Documentation here:
http://documentation.commvault.com/
Text-Only (commvault-traditional &lt; at &gt; yahoogroups.com?subject=Change Delivery Format: Traditional), Daily Digest (commvault-digest &lt; at &gt; yahoogroups.com?subject=Email Delivery: Digest) &bull; Unsubscribe (commvault-unsubscribe &lt; at &gt; yahoogroups.com?subject=Unsubscribe) &bull; Terms of Use

.


__,_._,___
Benchmarking Maglib Spindles?
February 15, 2012 05:15AM
Yes it&#8217;s been going all morning so far, with no obvious change in the response times reported.

Presumably I want to just leave it running if I want to see what it&#8217;s limitations are, or give it a rough number like 20000 to try and simulate how it would deal with 20tb of data?

From: commvault < at > yahoogroups.com [mailto] On Behalf Of Luke Walker
Sent: 15 February 2012 11:53
To: commvault < at > yahoogroups.com
Subject: RE: [commvault] Benchmarking Maglib Spindles?

Thanks, looking forward to the numbers.

Be warned it can take a while with no additional options as it&#8217;ll run as hard as it can until it hits its thresholds.. J

You can use the datasize switch (i.e. datasize 500, to simulate 0.5 TB of Application Data, or set it to something larger) to cut down the length of time it takes to run the test.

Cheers,
Luke

From: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com) [mailto] ([mailto]) On Behalf Of Paul Hutchings
Sent: Wednesday, 15 February 2012 9:17 PM
To: 'commvault < at > yahoogroups.com'
Subject: RE: [commvault] Benchmarking Maglib Spindles?

Thanks Luke, that&#8217;s been running for an hour and a half or so using SIDB2 -simulateddb -in instance001 -p e:dedup_store.

The SSD is an Intel 320. I&#8217;ll report back once I have some numbers.

From: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com) [mailto] ([mailto]) On Behalf Of Luke Walker
Sent: 14 February 2012 20:23
To: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?

Hey Paul,

You should be running a simulateddb from SIDB2 to get a more specific DDB benchmark:

http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/prod_info/dedup_disk.htm?var1=http://documentation.commvault.com/commvault/release_9! _0_0/books_online_1! /english_us/features/dedup_disk/tools.htm
Sample command:

SIDB2.exe simulateddb p Econfused smileyimulate in Instance001 datasize 2

Would be interested to see what your results are against the SSD J What model/manufacturer of SSD are you using.. Intel?

Cheers,
Luke

From: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com) [mailto] ([mailto]) On Behalf Of Paul Hutchings
Sent: Wednesday, 15 February 2012 3:36 AM
To: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?

For what it's wort! h, SSD is here, just used cvdiskperf to do a quick benchmark:

D iskPerf Version : 1.3
Path Used : Econfused smileyIDB01
Read-Write type : RANDOM
Block ! Size : 512
Block Count : 4096
File Count : 1024
Total Bytes Written : 2147483648
Time Taken to Write(S) : 9.89
Throughput Write(GB/H) : 727.82
Total Bytes Read : 2147483648
Time Taken to Read(S) : 8.56
Throughput Read(GB/H) : 841.13
Time Taken to Create(S) : 7.90
Throughput Create(GB/H) : 911.15

I haven't yet installed Commvault directly on the MA (I just copied the base folder) but if there's any more DDB specific benchmarks I'd be interested - the ones in BOL seemed to need the instancexxx to run?

Paul

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Paul Hutchings
Sent: 01 February 2012 21:24
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchma! rking Maglib Spindles?

See now I could lie and say I chose the Inte! l 320 because the documentations explicitly claims low write amplifica tion :-)

Truth is though, I chose it because it seemed to get decent reviews as a solid performer, and a couple of folks reported first hand using them an with a Dell PERC RAID controller without any issues.

When you say "after a while" with regard to WA, do you mean as in if I ran IOMeter you'd expect to see performance drop after minutes of sustained usage, or are you talking specifically long term? That's where the Wikipedia article started to get a little heavy.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 20! 12 9:12 PM
To: commvau lt < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?
! ;
I am a big fan of SSDs. Spinning disk performance has not kept up with the rest of the technologies like CPUs and memory. 10/15K drives are not going to survive in the long run...

Unfortunately, no detailed info was provided when we discussed the benchmarks. However, since the SIDB process can heavily load the device with a lot of random writes, and as much as TRIM does miracles to hide SSD write limitations (read/compare/write/verify), after a while you may run into issues like Write Amplification (http://en.wikipedia.org/wiki/Write_amplification) which ends up killing the performance down to a good old spinning disk... That's probably the point where you start crying ;-)

On 2012-02-01, at 3:56 PM, Brian Kelly wrote:

What SSD are you using ?
On Feb 1, 2012 3:52 PM, "! Paul Hutchings" <paul.hutchings < at > mira.c! o.uk (paul.hutchings%40mira.co.uk)> wrote:

I may be back crying like a baby smiling smiley I'm making some assumptions here but when you guys tested it what were you comparing it against?

We're firmly at the bottom end of the Commvault scale so if we had gone with RAID0 for the DDB we'd only have been able to stretch to 3 or 4.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 8:49 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

No problem. Let us know how you make out with the SSD. Curious to see how ! it behaves in the real world.

Regards,

Phil

On 2012-02-01, at 3:39 PM, Paul Hutchings wrote:

Thanks Phil, taking a ! good look at those. The DDB will be on an SSD. I recall in one of your replies to another of my posts you said along the line that SSD doesn't lend itself to DDB quite the same as you might expect, but from any/all benchmarks that I looked at it should hopefully still murder 3-4 10 or 15k spindles so I'm hopeful.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 8:10 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

Hi Paul,

You can use CVDISKPERF and the SIDB2 tools. CVDISK! PERF will help you benchmark your maglib while the SIDB2 tool will help you benchmark the disk(! s) on which you will put the DDB. By the way, disk performance for the DDB is a lot more critical than the one for the maglib itself. The faster your DDB disk(s) the faster your dedup will run.

For the docs, please see: http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/features/dedup_disk/tools.htm

Regards,

Phil

On 2012-02-01, at 2:50 PM, Paul Hutchings wrote:

What is the recommended way to benchmark various RAID configs for maglib volumes please?

Environment will be dedupe with auxcopy to tape.

I know I can use IOMeter, but I'm interested in any specific Commvault tools and what kind of profile (IO size, type, distribution etc.) I should use! .

Thanks,
Paul
--
MIRA Ltd

Watling Street, Nune aton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registr! ation GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

***************************Legal Disclaimer***************************
"This communication may contain confidential and privileged material for the
sole use of the intended recipient. Any unauthorized review, use or distribution
by others is strictly prohibited. If you have received the message in error,
please advise the sender by reply email and delete the message. Thank you."
**********************************************************************

MIRA Ltd

Watling Street, Nuneaton, Warwickshire, CV10 0TU, England

Registered in England and Wales No. 402570

VAT Registration GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

__._,_.___
Reply to sender (paul.hutchings &lt; at &gt; mira.co.uk?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply to group (commvault &lt; at &gt; yahoogroups.com?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply via web post | Start a New Topic
Messages in this topic (14)
Recent Activity:Visit Your Group

Commvault Documentation here:
http://documentation.commvault.com/
Text-Only (commvault-traditional &lt; at &gt; yahoogroups.com?subject=Change Delivery Format: Traditional), Daily Digest (commvault-digest &lt; at &gt; yahoogroups.com?subject=Email Delivery: Digest) &bull; Unsubscribe (commvault-unsubscribe &lt; at &gt; yahoogroups.com?subject=Unsubscribe) &bull; Terms of Use

.


__,_._,___
Benchmarking Maglib Spindles?
February 15, 2012 05:56AM
Here you go…

Ctongue sticking out smileyrogram FilesCommVaultSimpanaBase>SIDB2 -simulateddb -in instance001 -p e:dedup_store -datasize 1000

Creating new DB Files under [e]
Performing QueryInsert ... [Wed Feb 15 13]

[Parameters Used]
Application data size -> [1000] GB
Application data size per thread -> [100.00] GB
Dedupe Ratio -> [5]
Block Size -> [128] KB
No. of connections -> [10]
No. of records already present:
[0] Primary records, [0] Secondary records.

Iteration [810000] [Wed Feb 15 13]
Total QueryInsert time - [144.971869] secs
Total Commit time - [37.950870] secs
Average time for last [10000] operations:
QueryInsert - [123.38] microseconds
Commit - [14.99] microseconds
QueryInsert + Commit - [138.37] microseconds
Moving average - [0.00] microseconds

No. of records inserted upto threshold limit:
[1638410] Primary records, [8192000] Secondary records.

QueryInsert time taken per connection = [135.365391] secs
Max. QueryInsert time taken = [149.513710] secs
Commit time taken per connection = [41.102665] secs
Max. Commit time taken = [46.121138] secs
QueryInsert + Commit Time per connection = [176.468056] secs

Deduplication DB Simulation Completed [Wed Feb 15 13]

The disk is capable of hosting a Deduplication DB for:
0.977 TB of Application Data size
0.195 TB of data on disk
215.4 microseconds average Query & Insert time per block
Throughput for DDB server 20400 GB per Hour

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Luke Walker
Sent: 15 February 2012 11:53
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?

 
Thanks, looking forward to the numbers.
 
Be warned – it can take a while with no additional options as it’ll run as hard as it can until it hits its thresholds.. ☺
 
You can use the –datasize switch (i.e. –datasize 500, to simulate 0.5 TB of Application Data, or set it to something larger) to cut down the length of time it takes to run the test.
 
Cheers,
Luke
 
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Paul Hutchings
Sent: Wednesday, 15 February 2012 9:17 PM
To: 'commvault < at > yahoogroups.com (%27commvault%40yahoogroups.com)'
Subject: RE: [commvault] Benchmarking Maglib Spindles?
 
 
Thanks Luke, that’s been running for an hour and a half or so using “SIDB2 -simulateddb -in instance001 -p e:dedup_store”.
 
The SSD is an Intel 320.  I’ll report back once I have some numbers.
 
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Luke Walker
Sent: 14 February 2012 20:23
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?
 
 
Hey Paul,
 
You should be running a –simulateddb from SIDB2 to get a more specific DDB benchmark:
 
http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/prod_info/dedup_disk.htm?var1=http://documentation.commvault.com/commvault/release_9! _0_0/books_online_1! /english_us/features/dedup_disk/tools.htm
Sample command:
 
SIDB2.exe –simulateddb –p Econfused smileyimulate –in Instance001 –datasize 2
 
Would be interested to see what your results are against the SSD ☺   What model/manufacturer of SSD are you using.. Intel?
 
Cheers,
Luke
 
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Paul Hutchings
Sent: Wednesday, 15 February 2012 3:36 AM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?
 
 
For what it's wort! h, SSD is here, just used cvdiskperf to do a quick benchmark:

D iskPerf Version : 1.3
Path Used : Econfused smileyIDB01
Read-Write type : RANDOM
Block ! Size : 512
Block Count : 4096
File Count : 1024
Total Bytes Written : 2147483648
Time Taken to Write(S) : 9.89
Throughput Write(GB/H) : 727.82
Total Bytes Read : 2147483648
Time Taken to Read(S) : 8.56
Throughput Read(GB/H) : 841.13
Time Taken to Create(S) : 7.90
Throughput Create(GB/H) : 911.15

I haven't yet installed Commvault directly on the MA (I just copied the base folder) but if there's any more DDB specific benchmarks I'd be interested - the ones in BOL seemed to need the instancexxx to run?

Paul

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Paul Hutchings
Sent: 01 February 2012 21:24
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchma! rking Maglib Spindles?

 
See now I could lie and say I chose the Inte! l 320 because the documentations explicitly claims low write amplifica tion :-)

Truth is though, I chose it because it seemed to get decent reviews as a solid performer, and a couple of folks reported first hand using them an with a Dell PERC RAID controller without any issues.

When you say "after a while" with regard to WA, do you mean as in if I ran IOMeter you'd expect to see performance drop after minutes of sustained usage, or are you talking specifically long term?  That's where the Wikipedia article started to get a little heavy.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 20! 12 9:12 PM
To: commvau lt < at > yahoogroups.com (lt%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?
 ! ;
I am a big fan of SSDs. Spinning disk performance has not kept up with the rest of the technologies like CPUs and memory. 10/15K drives are not going to survive in the long run...

Unfortunately, no detailed info was provided when we discussed the benchmarks. However, since the SIDB process can heavily load the device with a lot of random writes, and as much as TRIM does miracles to hide SSD write limitations (read/compare/write/verify), after a while you may run into issues like Write Amplification (http://en.wikipedia.org/wiki/Write_amplification) which ends up killing the performance down to a good old spinning disk... That's probably the point where you start crying ;-)

On 2012-02-01, at 3:56 PM, Brian Kelly wrote:

 

What SSD are you using ?
On Feb 1, 2012 3:52 PM, "! Paul Hutchings" <paul.hutchings < at > mira.c (paul.hutchings%40mira.c)! o.uk> wrote:
 

I may be back crying like a baby smiling smiley  I'm making some assumptions here but when you guys tested it what were you comparing it against?

We're firmly at the bottom end of the Commvault scale so if we had gone with RAID0 for the DDB we'd only have been able to stretch to 3 or 4.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 8:49 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?
 
No problem. Let us know how you make out with the SSD. Curious to see how ! it behaves in the real world.

Regards, 

Phil

On 2012-02-01, at 3:39 PM, Paul Hutchings wrote:

 

Thanks Phil, taking a ! good look at those.  The DDB will be on an SSD.  I recall in one of your replies to another of my posts you said along the line that SSD doesn't lend itself to DDB quite the same as you might expect, but from any/all benchmarks that I looked at it should hopefully still murder 3-4 10 or 15k spindles so I'm hopeful.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 8:10 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?
 
Hi Paul, 

You can use CVDISKPERF and the SIDB2 tools. CVDISK! PERF will help you benchmark your maglib while the SIDB2 tool will help you benchmark the disk(! s) on which you will put the DDB. By the way, disk performance for the DDB is a lot more critical than the one for the maglib itself. The faster your DDB disk(s) the faster your dedup will run. 

For the docs, please see: http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/features/dedup_disk/tools.htm

Regards, 

Phil

On 2012-02-01, at 2:50 PM, Paul Hutchings wrote:

 
What is the recommended way to benchmark various RAID configs for maglib volumes please?

Environment will be dedupe with auxcopy to tape.

I know I can use IOMeter, but I'm interested in any specific Commvault tools and what kind of profile (IO size, type, distribution etc.) I should use! .

Thanks,
Paul
--
MIRA Ltd

Watling Street, Nune aton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registr! ation GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

***************************Legal Disclaimer***************************
"This communication may contain confidential and privileged material for the
sole use of the intended recipient. Any unauthorized review, use or distribution
by others is strictly prohibited. If you have received the message in error,
please advise the sender by reply email and delete the message. Thank you."
**********************************************************************
________________________________________
MIRA Ltd
 
Watling Street, Nuneaton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registration  GB 100 1464 84
 
The contents of this e-mail are confidential and are solely for the use of the intended recipient.  If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax.  You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

__._,_.___
Reply to sender (paul.hutchings &lt; at &gt; mira.co.uk?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply to group (commvault &lt; at &gt; yahoogroups.com?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply via web post | Start a New Topic
Messages in this topic (15)
Recent Activity:Visit Your Group

Commvault Documentation here:
http://documentation.commvault.com/
Text-Only (commvault-traditional &lt; at &gt; yahoogroups.com?subject=Change Delivery Format: Traditional), Daily Digest (commvault-digest &lt; at &gt; yahoogroups.com?subject=Email Delivery: Digest) &bull; Unsubscribe (commvault-unsubscribe &lt; at &gt; yahoogroups.com?subject=Unsubscribe) &bull; Terms of Use

.


__,_._,___
Benchmarking Maglib Spindles?
February 15, 2012 08:52PM
Thanks Paul!  Looks quite fast (average Query & Insert time per block / Throughput of DDB server)

Out of curiosity, what’s the hardware spec for your MA?

Cheers,
Luke

From: commvault < at > yahoogroups.com [mailto] On Behalf Of Paul Hutchings
Sent: Thursday, 16 February 2012 12:56 AM
To: 'commvault < at > yahoogroups.com'
Subject: RE: [commvault] Benchmarking Maglib Spindles?

Here you go…

Ctongue sticking out smileyrogram FilesCommVaultSimpanaBase>SIDB2 -simulateddb -in instance001 -p e:dedup_store -datasize 1000

Creating new DB Files under [e]
Performing QueryInsert ... [Wed Feb 15 13]

[Parameters Used]
Application data size -> [1000] GB
Application data size per thread -> [100.00] GB
Dedupe Ratio -> [5]
Block Size -> [128] KB
No. of connections -> [10]
No. of records already present:
[0] Primary records, [0] Secondary records.

Iteration [810000] [Wed Feb 15 13]
Total QueryInsert time - [144.971869] secs
Total Commit time - [37.950870] secs
Average time for last [10000] operations:
QueryInsert - [123.38] microseconds
Commit - [14.99] microseconds
QueryInsert + Commit - [138.37] microseconds
Moving average - [0.00] microseconds

No. of records inserted upto threshold limit:
[1638410] Primary records, [8192000] Secondary records.

QueryInsert time taken per connection = [135.365391] secs
Max. QueryInsert time taken = [149.513710] secs
Commit time taken per connection = [41.102665] secs
Max. Commit time taken = [46.121138] secs
QueryInsert + Commit Time per connection = [176.468056] secs

Deduplication DB Simulation Completed [Wed Feb 15 13]

The disk is capable of hosting a Deduplication DB for:
0.977 TB of Application Data size
0.195 TB of data on disk
215.4 microseconds average Query & Insert time per block
Throughput for DDB server 20400 GB per Hour

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Luke Walker
Sent: 15 February 2012 11:53
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?

Thanks, looking forward to the numbers.

Be warned – it can take a while with no additional options as it’ll run as hard as it can until it hits its thresholds.. ☺

You can use the –datasize switch (i.e. –datasize 500, to simulate 0.5 TB of Application Data, or set it to something larger) to cut down the length of time it takes to run the test.

Cheers,
Luke

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Paul Hutchings
Sent: Wednesday, 15 February 2012 9:17 PM
To: 'commvault < at > yahoogroups.com (%27commvault%40yahoogroups.com)'
Subject: RE: [commvault] Benchmarking Maglib Spindles?

Thanks Luke, that’s been running for an hour and a half or so using “SIDB2 -simulateddb -in instance001 -p e:dedup_store”.

The SSD is an Intel 320. I’ll report back once I have some numbers.

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Luke Walker
Sent: 14 February 2012 20:23
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?

Hey Paul,

You should be running a –simulateddb from SIDB2 to get a more specific DDB benchmark:

http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/prod_info/dedup_disk.htm?var1=http://documentation.commvault.com/commvault/release_9! _0_0/books_online_1! /english_us/features/dedup_disk/tools.htm
Sample command:

SIDB2.exe –simulateddb –p Econfused smileyimulate –in Instance001 –datasize 2

Would be interested to see what your results are against the SSD ☺ What model/manufacturer of SSD are you using.. Intel?

Cheers,
Luke

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Paul Hutchings
Sent: Wednesday, 15 February 2012 3:36 AM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?

For what it's wort! h, SSD is here, just used cvdiskperf to do a quick benchmark:

D iskPerf Version : 1.3
Path Used : Econfused smileyIDB01
Read-Write type : RANDOM
Block ! Size : 512
Block Count : 4096
File Count : 1024
Total Bytes Written : 2147483648
Time Taken to Write(S) : 9.89
Throughput Write(GB/H) : 727.82
Total Bytes Read : 2147483648
Time Taken to Read(S) : 8.56
Throughput Read(GB/H) : 841.13
Time Taken to Create(S) : 7.90
Throughput Create(GB/H) : 911.15

I haven't yet installed Commvault directly on the MA (I just copied the base folder) but if there's any more DDB specific benchmarks I'd be interested - the ones in BOL seemed to need the instancexxx to run?

Paul

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Paul Hutchings
Sent: 01 February 2012 21:24
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchma! rking Maglib Spindles?

See now I could lie and say I chose the Inte! l 320 because the documentations explicitly claims low write amplifica tion :-)

Truth is though, I chose it because it seemed to get decent reviews as a solid performer, and a couple of folks reported first hand using them an with a Dell PERC RAID controller without any issues.

When you say "after a while" with regard to WA, do you mean as in if I ran IOMeter you'd expect to see performance drop after minutes of sustained usage, or are you talking specifically long term? That's where the Wikipedia article started to get a little heavy.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 20! 12 9:12 PM
To: commvau lt < at > yahoogroups.com (lt%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?
! ;
I am a big fan of SSDs. Spinning disk performance has not kept up with the rest of the technologies like CPUs and memory. 10/15K drives are not going to survive in the long run...

Unfortunately, no detailed info was provided when we discussed the benchmarks. However, since the SIDB process can heavily load the device with a lot of random writes, and as much as TRIM does miracles to hide SSD write limitations (read/compare/write/verify), after a while you may run into issues like Write Amplification (http://en.wikipedia.org/wiki/Write_amplification) which ends up killing the performance down to a good old spinning disk... That's probably the point where you start crying ;-)

On 2012-02-01, at 3:56 PM, Brian Kelly wrote:

What SSD are you using ?
On Feb 1, 2012 3:52 PM, "! Paul Hutchings" <paul.hutchings < at > mira.c (paul.hutchings%40mira.c)! o.uk> wrote:

I may be back crying like a baby smiling smiley I'm making some assumptions here but when you guys tested it what were you comparing it against?

We're firmly at the bottom end of the Commvault scale so if we had gone with RAID0 for the DDB we'd only have been able to stretch to 3 or 4.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 8:49 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

No problem. Let us know how you make out with the SSD. Curious to see how ! it behaves in the real world.

Regards,

Phil

On 2012-02-01, at 3:39 PM, Paul Hutchings wrote:

Thanks Phil, taking a ! good look at those. The DDB will be on an SSD. I recall in one of your replies to another of my posts you said along the line that SSD doesn't lend itself to DDB quite the same as you might expect, but from any/all benchmarks that I looked at it should hopefully still murder 3-4 10 or 15k spindles so I'm hopeful.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 8:10 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

Hi Paul,

You can use CVDISKPERF and the SIDB2 tools. CVDISK! PERF will help you benchmark your maglib while the SIDB2 tool will help you benchmark the disk(! s) on which you will put the DDB. By the way, disk performance for the DDB is a lot more critical than the one for the maglib itself. The faster your DDB disk(s) the faster your dedup will run.

For the docs, please see: http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/features/dedup_disk/tools.htm

Regards,

Phil

On 2012-02-01, at 2:50 PM, Paul Hutchings wrote:

What is the recommended way to benchmark various RAID configs for maglib volumes please?

Environment will be dedupe with auxcopy to tape.

I know I can use IOMeter, but I'm interested in any specific Commvault tools and what kind of profile (IO size, type, distribution etc.) I should use! .

Thanks,
Paul
--
MIRA Ltd

Watling Street, Nune aton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registr! ation GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

***************************Legal Disclaimer***************************
"This communication may contain confidential and privileged material for the
sole use of the intended recipient. Any unauthorized review, use or distribution
by others is strictly prohibited. If you have received the message in error,
please advise the sender by reply email and delete the message. Thank you."
**********************************************************************
________________________________________
MIRA Ltd

Watling Street, Nuneaton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registration GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

__._,_.___
Reply to sender (luke &lt; at &gt; blackduck.nu?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply to group (commvault &lt; at &gt; yahoogroups.com?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply via web post | Start a New Topic
Messages in this topic (16)
Recent Activity:Visit Your Group

Commvault Documentation here:
http://documentation.commvault.com/
Text-Only (commvault-traditional &lt; at &gt; yahoogroups.com?subject=Change Delivery Format: Traditional), Daily Digest (commvault-digest &lt; at &gt; yahoogroups.com?subject=Email Delivery: Digest) &bull; Unsubscribe (commvault-unsubscribe &lt; at &gt; yahoogroups.com?subject=Unsubscribe) &bull; Terms of Use

.


__,_._,___
Benchmarking Maglib Spindles?
February 15, 2012 09:50PM
Interesting. I wonder how the numbers would look like with fusionIO or the new EMC VFcache cards! winking smiley. Great commands. Thx.

Paul M.

Sent from my iPad

On Feb 15, 2012, at 7:23 AM, "Luke Walker" <luke < at > blackduck.nu (luke &lt; at &gt; blackduck.nu)> wrote:

Quote


Thanks Paul! Looks quite fast (average Query & Insert time per block / Throughput of DDB server)

Out of curiosity, what’s the hardware spec for your MA?

Cheers,
Luke

From: commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com) [mailto] On Behalf Of Paul Hutchings
Sent: Thursday, 16 February 2012 12:56 AM
To: 'commvault < at > yahoogroups.com (commvault &lt; at &gt; yahoogroups.com)'
Subject: RE: [commvault] Benchmarking Maglib Spindles?

Here you go…

Ctongue sticking out smileyrogram FilesCommVaultSimpanaBase>SIDB2 -simulateddb -in instance001 -p e:dedup_store -datasize 1000

Creating new DB Files under [e]
Performing QueryInsert ... [Wed Feb 15 13]

[Parameters Used]
Application data size -> [1000] GB
Application data size per thread -> [100.00] GB
Dedupe Ratio -> [5]
Block Size -> [128] KB
No. of connections -> [10]
No. of records already present:
[0] Primary records, [0] Secondary records.

Iteration [810000] [Wed Feb 15 13]
Total QueryInsert time - [144.971869] secs
Total Commit time - [37.950870] secs
Average time for last [10000] operations:
QueryInsert - [123.38] microseconds
Commit - [14.99] microseconds
QueryInsert + Commit - [138.37] microseconds
Moving average - [0.00] microseconds

No. of records inserted upto threshold limit:
[1638410] Primary records, [8192000] Secondary records.

QueryInsert time taken per connection = [135.365391] secs
Max. QueryInsert time taken = [149.513710] secs
Commit time taken per connection = [41.102665] secs
Max. Commit time taken = [46.121138] secs
QueryInsert + Commit Time per connection = [176.468056] secs

Deduplication DB Simulation Completed [Wed Feb 15 13]

The disk is capable of hosting a Deduplication DB for:
0.977 TB of Application Data size
0.195 TB of data on disk
215.4 microseconds average Query & Insert time per block
Throughput for DDB server 20400 GB per Hour

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault < at > yahoogroups.com (commvault%40yahoogroups.com)[/url]] On Behalf Of Luke Walker
Sent: 15 February 2012 11:53
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?

Thanks, looking forward to the numbers.

Be warned – it can take a while with no additional options as it’ll run as hard as it can until it hits its thresholds.. ☺

You can use the –datasize switch (i.e. –datasize 500, to simulate 0.5 TB of Application Data, or set it to something larger) to cut down the length of time it takes to run the test.

Cheers,
Luke

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault < at > yahoogroups.com (commvault%40yahoogroups.com)[/url]] On Behalf Of Paul Hutchings
Sent: Wednesday, 15 February 2012 9:17 PM
To: 'commvault < at > yahoogroups.com (%27commvault%40yahoogroups.com)'
Subject: RE: [commvault] Benchmarking Maglib Spindles?

Thanks Luke, that’s been running for an hour and a half or so using “SIDB2 -simulateddb -in instance001 -p e:dedup_store”.

The SSD is an Intel 320. I’ll report back once I have some numbers.

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault < at > yahoogroups.com (commvault%40yahoogroups.com)[/url]] On Behalf Of Luke Walker
Sent: 14 February 2012 20:23
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?

Hey Paul,

You should be running a –simulateddb from SIDB2 to get a more specific DDB benchmark:

http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/prod_info/dedup_disk.htm?var1=http://documentation.commvault.com/commvault/release_9! _0_0/books_online_1! /english_us/features/dedup_disk/tools.htm
Sample command:

SIDB2.exe –simulateddb –p Econfused smileyimulate –in Instance001 –datasize 2

Would be interested to see what your results are against the SSD ☺ What model/manufacturer of SSD are you using.. Intel?

Cheers,
Luke

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault < at > yahoogroups.com (commvault%40yahoogroups.com)[/url]] On Behalf Of Paul Hutchings
Sent: Wednesday, 15 February 2012 3:36 AM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?

For what it's wort! h, SSD is here, just used cvdiskperf to do a quick benchmark:

D iskPerf Version : 1.3
Path Used : Econfused smileyIDB01
Read-Write type : RANDOM
Block ! Size : 512
Block Count : 4096
File Count : 1024
Total Bytes Written : 2147483648
Time Taken to Write(S) : 9.89
Throughput Write(GB/H) : 727.82
Total Bytes Read : 2147483648
Time Taken to Read(S) : 8.56
Throughput Read(GB/H) : 841.13
Time Taken to Create(S) : 7.90
Throughput Create(GB/H) : 911.15

I haven't yet installed Commvault directly on the MA (I just copied the base folder) but if there's any more DDB specific benchmarks I'd be interested - the ones in BOL seemed to need the instancexxx to run?

Paul

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault < at > yahoogroups.com (commvault%40yahoogroups.com)[/url]] On Behalf Of Paul Hutchings
Sent: 01 February 2012 21:24
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchma! rking Maglib Spindles?

See now I could lie and say I chose the Inte! l 320 because the documentations explicitly claims low write amplifica tion :-)

Truth is though, I chose it because it seemed to get decent reviews as a solid performer, and a couple of folks reported first hand using them an with a Dell PERC RAID controller without any issues.

When you say "after a while" with regard to WA, do you mean as in if I ran IOMeter you'd expect to see performance drop after minutes of sustained usage, or are you talking specifically long term? That's where the Wikipedia article started to get a little heavy.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [[url=mailto]commvault < at > yahoogroups.com (commvault%40yahoogroups.com)[/url]] on behalf of Philippe Normand [[url=mailto]pnormand < at > commvault.com (pnormand%40commvault.com)[/url]]
Sent: 01 February 20! 12 9:12 PM
To: commvau lt < at > yahoogroups.com (lt%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?
! ;
I am a big fan of SSDs. Spinning disk performance has not kept up with the rest of the technologies like CPUs and memory. 10/15K drives are not going to survive in the long run...

Unfortunately, no detailed info was provided when we discussed the benchmarks. However, since the SIDB process can heavily load the device with a lot of random writes, and as much as TRIM does miracles to hide SSD write limitations (read/compare/write/verify), after a while you may run into issues like Write Amplification (http://en.wikipedia.org/wiki/Write_amplification) which ends up killing the performance down to a good old spinning disk... That's probably the point where you start crying ;-)

On 2012-02-01, at 3:56 PM, Brian Kelly wrote:

What SSD are you using ?
On Feb 1, 2012 3:52 PM, "! Paul Hutchings" <paul.hutchings < at > mira.c (paul.hutchings%40mira.c)! o.uk> wrote:

I may be back crying like a baby smiling smiley I'm making some assumptions here but when you guys tested it what were you comparing it against?

We're firmly at the bottom end of the Commvault scale so if we had gone with RAID0 for the DDB we'd only have been able to stretch to 3 or 4.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [[url=mailto]commvault < at > yahoogroups.com (commvault%40yahoogroups.com)[/url]] on behalf of Philippe Normand [[url=mailto]pnormand < at > commvault.com (pnormand%40commvault.com)[/url]]
Sent: 01 February 2012 8:49 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

No problem. Let us know how you make out with the SSD. Curious to see how ! it behaves in the real world.

Regards,

Phil

On 2012-02-01, at 3:39 PM, Paul Hutchings wrote:

Thanks Phil, taking a ! good look at those. The DDB will be on an SSD. I recall in one of your replies to another of my posts you said along the line that SSD doesn't lend itself to DDB quite the same as you might expect, but from any/all benchmarks that I looked at it should hopefully still murder 3-4 10 or 15k spindles so I'm hopeful.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [[url=mailto]commvault < at > yahoogroups.com (commvault%40yahoogroups.com)[/url]] on behalf of Philippe Normand [[url=mailto]pnormand < at > commvault.com (pnormand%40commvault.com)[/url]]
Sent: 01 February 2012 8:10 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?

Hi Paul,

You can use CVDISKPERF and the SIDB2 tools. CVDISK! PERF will help you benchmark your maglib while the SIDB2 tool will help you benchmark the disk(! s) on which you will put the DDB. By the way, disk performance for the DDB is a lot more critical than the one for the maglib itself. The faster your DDB disk(s) the faster your dedup will run.

For the docs, please see: http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/features/dedup_disk/tools.htm

Regards,

Phil

On 2012-02-01, at 2:50 PM, Paul Hutchings wrote:

What is the recommended way to benchmark various RAID configs for maglib volumes please?

Environment will be dedupe with auxcopy to tape.

I know I can use IOMeter, but I'm interested in any specific Commvault tools and what kind of profile (IO size, type, distribution etc.) I should use! .

Thanks,
Paul
--
MIRA Ltd

Watling Street, Nune aton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registr! ation GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

***************************Legal Disclaimer***************************
"This communication may contain confidential and privileged material for the
sole use of the intended recipient. Any unauthorized review, use or distribution
by others is strictly prohibited. If you have received the message in error,
please advise the sender by reply email and delete the message. Thank you."
**********************************************************************
________________________________________
MIRA Ltd

Watling Street, Nuneaton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registration GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.
__._,_.___
Reply to sender (paul.montoya &lt; at &gt; gmail.com?subject=Re%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply to group (commvault &lt; at &gt; yahoogroups.com?subject=Re%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply via web post | Start a New Topic
Messages in this topic (17)
Recent Activity:Visit Your Group

Commvault Documentation here:
http://documentation.commvault.com/
Text-Only (commvault-traditional &lt; at &gt; yahoogroups.com?subject=Change Delivery Format: Traditional), Daily Digest (commvault-digest &lt; at &gt; yahoogroups.com?subject=Email Delivery: Digest) &bull; Unsubscribe (commvault-unsubscribe &lt; at &gt; yahoogroups.com?subject=Unsubscribe) &bull; Terms of Use

.


__,_._,___
Benchmarking Maglib Spindles?
February 16, 2012 12:55AM
Luke, I ran it again with a 20tb dataset, numbers below.

The MA is a Dell R710 with a pair of E5645's and 32gb of RAM. The DDB is the Intel SSD on a PERC H700.

Ctongue sticking out smileyrogram FilesCommVaultSimpanaBase>SIDB2 -simulateddb -in instance001 -p e:dedup_store -datasize 20000

Creating new DB Files under [e]
Performing QueryInsert ... [Wed Feb 15 14]

[Parameters Used]
Application data size -> [20000] GB
Application data size per thread -> [2000.00] GB
Dedupe Ratio -> [5]
Block Size -> [128] KB
No. of connections -> [10]
No. of records already present:
[0] Primary records, [0] Secondary records.

Iteration [16380000] [Wed Feb 15 15]
Total QueryInsert time - [2697.598337] secs
Total Commit time - [1249.212524] secs
Average time for last [10000] operations:
QueryInsert - [72.90] microseconds
Commit - [9.90] microseconds
QueryInsert + Commit - [82.80] microseconds
Moving average - [0.00] microseconds

No. of records inserted upto threshold limit:
[32768010] Primary records, [163840000] Secondary records.

QueryInsert time taken per connection = [2527.270930] secs
Max. QueryInsert time taken = [2731.912354] secs
Commit time taken per connection = [1331.092923] secs
Max. Commit time taken = [1384.717382] secs
QueryInsert + Commit Time per connection = [3858.363853] secs

Deduplication DB Simulation Completed [Wed Feb 15 15]

The disk is capable of hosting a Deduplication DB for:
19.531 TB of Application Data size
3.906 TB of data on disk
235.5 microseconds average Query & Insert time per block
Throughput for DDB server 18661 GB per Hour

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Luke Walker
Sent: 15 February 2012 14:23
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?

 
Thanks Paul!  Looks quite fast (average Query & Insert time per block / Throughput of DDB server)
 
Out of curiosity, what’s the hardware spec for your MA?
 
Cheers,
Luke
 
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Paul Hutchings
Sent: Thursday, 16 February 2012 12:56 AM
To: 'commvault < at > yahoogroups.com (%27commvault%40yahoogroups.com)'
Subject: RE: [commvault] Benchmarking Maglib Spindles?
 
 
Here you go…

Ctongue sticking out smileyrogram FilesCommVaultSimpanaBase>SIDB2 -simulateddb -in instance001 -p e:dedup_store -datasize 1000

Creating new DB Files under [e]
Performing QueryInsert ... [Wed Feb 15 13]

[Parameters Used]
Application data size -> [1000] GB
Application data size per thread -> [100.00] GB
Dedupe Ratio -> [5]
Block Size -> [128] KB
No. of connections -> [10]
No. of records already present:
[0] Primary records, [0] Secondary records.

Iteration [810000] [Wed Feb 15 13]
Total QueryInsert time - [144.971869] secs
Total Commit time - [37.950870] secs
Average time for last [10000] operations:
QueryInsert - [123.38] microseconds
Commit - [14.99] microseconds
QueryInsert + Commit - [138.37] microseconds
Moving average - [0.00] microseconds

No. of records inserted upto threshold limit:
[1638410] Primary records, [8192000] Secondary records.

QueryInsert time taken per connection = [135.365391] secs
Max. QueryInsert time taken = [149.513710] secs
Commit time taken per connection = [41.102665] secs
Max. Commit time taken = [46.121138] secs
QueryInsert + Commit Time per connection = [176.468056] secs

Deduplication DB Simulation Completed [Wed Feb 15 13]

The disk is capable of hosting a Deduplication DB for:
0.977 TB of Application Data size
0.195 TB of data on disk
215.4 microseconds average Query & Insert time per block
Throughput for DDB server 20400 GB per Hour

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Luke Walker
Sent: 15 February 2012 11:53
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?

 
Thanks, looking forward to the numbers.
 
Be warned – it can take a while with no additional options as it’ll run as hard as it can until it hits its thresholds.. ☺
 
You can use the –datasize switch (i.e. –datasize 500, to simulate 0.5 TB of Application Data, or set it to something larger) to cut down the length of time it takes to run the test.
 
Cheers,
Luke
 
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Paul Hutchings
Sent: Wednesday, 15 February 2012 9:17 PM
To: 'commvault < at > yahoogroups.com (%27commvault%40yahoogroups.com)'
Subject: RE: [commvault] Benchmarking Maglib Spindles?
 
 
Thanks Luke, that’s been running for an hour and a half or so using “SIDB2 -simulateddb -in instance001 -p e:dedup_store”.
 
The SSD is an Intel 320.  I’ll report back once I have some numbers.
 
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Luke Walker
Sent: 14 February 2012 20:23
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?
 
 
Hey Paul,
 
You should be running a –simulateddb from SIDB2 to get a more specific DDB benchmark:
 
http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/prod_info/dedup_disk.htm?var1=http://documentation.commvault.com/commvault/release_9! _0_0/books_online_1! /english_us/features/dedup_disk/tools.htm
Sample command:
 
SIDB2.exe –simulateddb –p Econfused smileyimulate –in Instance001 –datasize 2
 
Would be interested to see what your results are against the SSD ☺   What model/manufacturer of SSD are you using.. Intel?
 
Cheers,
Luke
 
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Paul Hutchings
Sent: Wednesday, 15 February 2012 3:36 AM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchmarking Maglib Spindles?
 
 
For what it's wort! h, SSD is here, just used cvdiskperf to do a quick benchmark:

D iskPerf Version : 1.3
Path Used : Econfused smileyIDB01
Read-Write type : RANDOM
Block ! Size : 512
Block Count : 4096
File Count : 1024
Total Bytes Written : 2147483648
Time Taken to Write(S) : 9.89
Throughput Write(GB/H) : 727.82
Total Bytes Read : 2147483648
Time Taken to Read(S) : 8.56
Throughput Read(GB/H) : 841.13
Time Taken to Create(S) : 7.90
Throughput Create(GB/H) : 911.15

I haven't yet installed Commvault directly on the MA (I just copied the base folder) but if there's any more DDB specific benchmarks I'd be interested - the ones in BOL seemed to need the instancexxx to run?

Paul

From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [mailto]commvault%40yahoogroups.com[/email])] On Behalf Of Paul Hutchings
Sent: 01 February 2012 21:24
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: RE: [commvault] Benchma! rking Maglib Spindles?

 
See now I could lie and say I chose the Inte! l 320 because the documentations explicitly claims low write amplifica tion :-)

Truth is though, I chose it because it seemed to get decent reviews as a solid performer, and a couple of folks reported first hand using them an with a Dell PERC RAID controller without any issues.

When you say "after a while" with regard to WA, do you mean as in if I ran IOMeter you'd expect to see performance drop after minutes of sustained usage, or are you talking specifically long term?  That's where the Wikipedia article started to get a little heavy.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 20! 12 9:12 PM
To: commvau lt < at > yahoogroups.com (lt%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?
 ! ;
I am a big fan of SSDs. Spinning disk performance has not kept up with the rest of the technologies like CPUs and memory. 10/15K drives are not going to survive in the long run...

Unfortunately, no detailed info was provided when we discussed the benchmarks. However, since the SIDB process can heavily load the device with a lot of random writes, and as much as TRIM does miracles to hide SSD write limitations (read/compare/write/verify), after a while you may run into issues like Write Amplification (http://en.wikipedia.org/wiki/Write_amplification) which ends up killing the performance down to a good old spinning disk... That's probably the point where you start crying ;-)

On 2012-02-01, at 3:56 PM, Brian Kelly wrote:

 

What SSD are you using ?
On Feb 1, 2012 3:52 PM, "! Paul Hutchings" <paul.hutchings < at > mira.c (paul.hutchings%40mira.c)! o.uk> wrote:
 

I may be back crying like a baby smiling smiley  I'm making some assumptions here but when you guys tested it what were you comparing it against?

We're firmly at the bottom end of the Commvault scale so if we had gone with RAID0 for the DDB we'd only have been able to stretch to 3 or 4.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 8:49 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?
 
No problem. Let us know how you make out with the SSD. Curious to see how ! it behaves in the real world.

Regards, 

Phil

On 2012-02-01, at 3:39 PM, Paul Hutchings wrote:

 

Thanks Phil, taking a ! good look at those.  The DDB will be on an SSD.  I recall in one of your replies to another of my posts you said along the line that SSD doesn't lend itself to DDB quite the same as you might expect, but from any/all benchmarks that I looked at it should hopefully still murder 3-4 10 or 15k spindles so I'm hopeful.
________________________________________
From: commvault < at > yahoogroups.com (commvault%40yahoogroups.com) [commvault < at > yahoogroups.com (commvault%40yahoogroups.com)] on behalf of Philippe Normand [pnormand < at > commvault.com (pnormand%40commvault.com)]
Sent: 01 February 2012 8:10 PM
To: commvault < at > yahoogroups.com (commvault%40yahoogroups.com)
Subject: Re: [commvault] Benchmarking Maglib Spindles?
 
Hi Paul, 

You can use CVDISKPERF and the SIDB2 tools. CVDISK! PERF will help you benchmark your maglib while the SIDB2 tool will help you benchmark the disk(! s) on which you will put the DDB. By the way, disk performance for the DDB is a lot more critical than the one for the maglib itself. The faster your DDB disk(s) the faster your dedup will run. 

For the docs, please see: http://documentation.commvault.com/commvault/release_9_0_0/books_online_1/english_us/features/dedup_disk/tools.htm

Regards, 

Phil

On 2012-02-01, at 2:50 PM, Paul Hutchings wrote:

 
What is the recommended way to benchmark various RAID configs for maglib volumes please?

Environment will be dedupe with auxcopy to tape.

I know I can use IOMeter, but I'm interested in any specific Commvault tools and what kind of profile (IO size, type, distribution etc.) I should use! .

Thanks,
Paul
--
MIRA Ltd

Watling Street, Nune aton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registr! ation GB 100 1464 84

The contents of this e-mail are confidential and are solely for the use of the intended recipient. If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax. You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

***************************Legal Disclaimer***************************
"This communication may contain confidential and privileged material for the
sole use of the intended recipient. Any unauthorized review, use or distribution
by others is strictly prohibited. If you have received the message in error,
please advise the sender by reply email and delete the message. Thank you."
**********************************************************************
________________________________________
MIRA Ltd
 
Watling Street, Nuneaton, Warwickshire, CV10 0TU, England
Registered in England and Wales No. 402570
VAT Registration  GB 100 1464 84
 
The contents of this e-mail are confidential and are solely for the use of the intended recipient.  If you receive this e-mail in error, please delete it and notify us either by e-mail, telephone or fax.  You should not copy, forward or otherwise disclose the content of the e-mail as this is prohibited.

__._,_.___
Reply to sender (paul.hutchings &lt; at &gt; mira.co.uk?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply to group (commvault &lt; at &gt; yahoogroups.com?subject=RE%3A%20%5Bcommvault%5D%20Benchmarking%20Maglib%20Spindles%3F) | Reply via web post | Start a New Topic
Messages in this topic (18)
Recent Activity:Visit Your Group

Commvault Documentation here:
http://documentation.commvault.com/
Text-Only (commvault-traditional &lt; at &gt; yahoogroups.com?subject=Change Delivery Format: Traditional), Daily Digest (commvault-digest &lt; at &gt; yahoogroups.com?subject=Email Delivery: Digest) &bull; Unsubscribe (commvault-unsubscribe &lt; at &gt; yahoogroups.com?subject=Unsubscribe) &bull; Terms of Use

.


__,_._,___
Sorry, only registered users may post in this forum.

Click here to login