cancel
Showing results for 
Search instead for 
Did you mean: 

PureDisk performance issue

Amaan
Level 6

Hi everyone,

I am working on the performance issue of the duplciation jobs from PD.

I found in the following TN command which will test read and write test:

http://www.symantec.com/business/support/index?page=content&id=TECH156743

Command syntax is:

time (dd if=/dev/zero of=/Storage/data/xyz bs=64k count=409600; sync)

When i am runnig this command i am getting the results as mentioned in the TN. I have to make sure that i am getting proper results.

Actually i am new in PD and dont know well about it.

Is this command identical for all PD tests? if no what should i change?

Is there any other tests which i can check/test to find the root cause of slow duplications. What kind of test i can do to check network speed?

Thanks!

1 ACCEPTED SOLUTION

Accepted Solutions

Amaan
Level 6

I have reviewed lots of staff regarding this and your comments helped alot on this. as the question was about the PD performance issue and if the command syntax is correct, I cannot mark anybodies post as solution as they have been just guidence not an answer. i cannot mark current post as solution as well, but it has some more info which could help others.

I am going to share this to guide anybody else who will face with same issue or similar issue

I worked with Symantec on this and we have done following to test performance issue of the PD:

  1. Media server performance (available memory, CPU and etc..);
  2. Media server NIC througput (Network performance); - just did some monitoring on the media server to check if we have enough throuput on media server NIC.
  3. PDOS performance; - this has been checked by Symantec engineer. but wrote it down for future purposes. here is how to check it:

In the /Storage/log/spoold/storaged.log find two entries.

1.  November 05 11:54:08 INFO [1077967168]: Synchronization for transaction   log /Storage/queue/sorted-494123-495860.tlog started, 936016 transactions pending..

2. November 05 13:09:03 INFO [1077967168]: Time required to build index on objects2 table: xxxxxx.xxxxx

3.  Subtract the time example  13:09:02 - 11:54:08 = 75 (minutes in my example)

4.  Divide the number of transactions by the number of minutes
     936016 / 75 = 12480.2133 transactions per minute
My transactions per minute is:
   13:26:11 - 10:10:24 = 196minutes
   34193021 / 196 = 174454.1887 transactions per minute

Anything above 100,000 transactions per minute is good.

       4. PD disk performance; will be checked in two ways. by running dd command or by camel tool. you can find more details here:

http://www.symantec.com/business/support/index?page=content&id=TECH156743

      Issue with our PD was on disk performance and NIC. NIC on media server was highly utilised and PD disk performance is ~60MB, recommanded is to be more than 150MB.

And whoever have NBU ver 7.1.0.3 and more has to do some tuning as well as mentioned in Mariannes post: NetBackup 7.1.0.3 rehydration improvements
 

Thanks everyone for your answers!

 

 

 

View solution in original post

6 REPLIES 6

Amaan
Level 6
Please need help ASAP. 35 is ready. :)

Amaan
Level 6
I dont think that nobody knows answers. Please help. I cannot open case for this performance issue.

teiva-boy
Level 6

There is a known issue with duplication to tape and performance.  Is that what you are experiencing?  

EItherway, I don't understand your question.  You're running the dd commands as specified and you are getting the resultant values, which I think are acceptable according to the TN? (>200MB/s?)

The TN then goes further to say call support for the camel tool or VxFS tools.  

There are a ton of other ways to test file copy speeds over the network using ftp, dd, and many others that are native to linux.

 

You should be able to open a case with support over performance issues.  They can identify those issues by validating your results, or referring you to the known issue of duplications to tape.  When in doubt with support, ASK FOR A DUTY MANAGER.

Amaan
Level 6
Thanks for your reply. My question was to know if this command is comman for everyone which has a PD. And result was 50-70MB. Thats the reason i asked this question. :) One more question. Is this command read and write test, or it will just tests write speed of PD? I am not good in unix systems and commands like dd. Will be waiting for your answer. Thanks, I will try to open case with symantec.

Marianne
Level 6
Partner    VIP    Accredited Certified

Have a look at the Release Notes for 7.1.0.3http://symantec.com/docs/DOC4886

Appendix A contains the following:

NetBackup 7.1.0.3 rehydration improvements
This appendix includes the following topics:
■ About the NetBackup 7.1.0.3 rehydration improvements for deduplicated files
■ Environmental factors that affect rehydration performance
■ Changes and updates to the deduplication tuning parameters that affect restore performance and rehydration performance
■ Editing the ReadBuffer Size parameter
■ NetBackup tuning parameters that affect restore performance and rehydration performance

Amaan
Level 6

I have reviewed lots of staff regarding this and your comments helped alot on this. as the question was about the PD performance issue and if the command syntax is correct, I cannot mark anybodies post as solution as they have been just guidence not an answer. i cannot mark current post as solution as well, but it has some more info which could help others.

I am going to share this to guide anybody else who will face with same issue or similar issue

I worked with Symantec on this and we have done following to test performance issue of the PD:

  1. Media server performance (available memory, CPU and etc..);
  2. Media server NIC througput (Network performance); - just did some monitoring on the media server to check if we have enough throuput on media server NIC.
  3. PDOS performance; - this has been checked by Symantec engineer. but wrote it down for future purposes. here is how to check it:

In the /Storage/log/spoold/storaged.log find two entries.

1.  November 05 11:54:08 INFO [1077967168]: Synchronization for transaction   log /Storage/queue/sorted-494123-495860.tlog started, 936016 transactions pending..

2. November 05 13:09:03 INFO [1077967168]: Time required to build index on objects2 table: xxxxxx.xxxxx

3.  Subtract the time example  13:09:02 - 11:54:08 = 75 (minutes in my example)

4.  Divide the number of transactions by the number of minutes
     936016 / 75 = 12480.2133 transactions per minute
My transactions per minute is:
   13:26:11 - 10:10:24 = 196minutes
   34193021 / 196 = 174454.1887 transactions per minute

Anything above 100,000 transactions per minute is good.

       4. PD disk performance; will be checked in two ways. by running dd command or by camel tool. you can find more details here:

http://www.symantec.com/business/support/index?page=content&id=TECH156743

      Issue with our PD was on disk performance and NIC. NIC on media server was highly utilised and PD disk performance is ~60MB, recommanded is to be more than 150MB.

And whoever have NBU ver 7.1.0.3 and more has to do some tuning as well as mentioned in Mariannes post: NetBackup 7.1.0.3 rehydration improvements
 

Thanks everyone for your answers!