Best settings for Data Protector 9.0 and LTO-6 drive

[P_Buchanan]

Greetings,

I am looking for what the Best settings for Data Protector 9.0 and LTO-6 drive might be. Block size, Segment size,  Disk agent buffers, etc.

Pat Buchanan

pbucha@iscarusa.com

[svollrat]

We are currently using 256k Blocks, 10'000MB Segments and 32 DA-Buffers with Concurrency 16, works well so far.

You can reduce the DA-Buffers if your clients dan deliver sufficient Data and if fast single file Restore is not too important you can increase the Segment-Size to 20'000 or more.

[P_Buchanan]

As an additional note,  we are going straight from the Storeeasy's to tape once a week. No clients are backing up to tape directly. It's a scheduled object copy once per week.

[jruffer]

We are still on 7.03 but using LTO6.  Backups are to StoreOnce Catalyst sessions are copied to tape.

We use a Block size of 512. Segment size 20000, Disk agent buffers 8.  Concurrency of 1.

We were told that clients would need reconfiguration to allow a block size of 1024 which would have been better but was not practical at the time.

HTH

Jeremy

[P_Buchanan]

Do you have any gauge as to time? I have one drive, and it takes me about 27 hours to get 6.5 tb onto 2 tapes.

[P_Buchanan]

Do you have any gauge as to time as well? I have one drive, and it takes me about 27 hours to get 6.5 tb onto 2 tapes.

[Bob_Clark]

The one thing that was not mentioned, and that may really slow down the Backup is Tape Drive Streaming.  This can cut eorformance by up to 90% if the drive needs to keep stopping/starting

We usually address this by incrweasing the concurrency (number of Data Streams) to a value of 10 or 12.  Of course, this requires that you have enough data streams/Objects/mounted file systms flowing to the drive at the same time.  If you are only backing up 2 or 3 very large Objects, then setting concurrency higher is probably not going to help you

Please review the excellent PowerPoint presentation one of my colleagues wrte about this (attached).  I have also attached a Performance WhitePaper

[P_Buchanan]

Quite a bit of great information Bob, thanks!

I did some of the checks on my last tape session. Performance was 230gb/hour. I had 209 objects and 87 Completed Media Agents. I am setup with 512kb block sizes, both my concurrency and disk agent buffers are set at 32. I calculated my segment size based on the help menu to 52,429. My media server has 32gb. I wrote 6.23 tb onto two LTO-6 tapes in 27-hours. I hit a rate as high as 540mb/minute.

Not sure if I can increase performance any more than this, but will continue to test and study. Thanks for the help.

[svollrat]

the official Rates for LTO6 are 160MB/s uncompressed and 400MB/s compressed so you probably won't exceed your measured peak anymore. Guess only thing you can tinker around with is increasing the average Speed and especialy the lower ends that tend to lead to start/stop operations.

Giving a good general recommendation for that is hard to impossible. Well besides preventing one slow stream to run on its own for longer period of time of course 😉