TSM Diskpool

[spiffy]

Wondering what everyone is using for their TSM diskpool.
Currently I have 200GB diskpool, on local storage on the TSM server.
The SAN guys have allocated 4TB of disk on a Symmetrix, which i am going to split between my 2 tsm servers, but am wondering how large i should make the diskpool volumes and how many i should have.
I "could" just allocate 1 large diskpool volume, but that wouldnt be smart, so i will probably cut it up into a bunch of volumes...

So wondering if anyone else is using SAN storage as their Diskpool, if you have a dedicated HBA for the diskpool, and how you cut up your volumes.

Thanks
James

 

[THE_WIPET]

ON our 40TB diskpool(it was on a DS4100) and I had 3 TSM server accessing it. (it a w2k3 x64 ENV).

The arrays were created to take the maximum spindle on the SAN. Each storage pool had there LUN. Some of the disk were span on 2 or 3 LUN of 1.3TB) On each disk we created in TSM multiple volumes of 200gb.the biggest diskpool for a storage group was 4TB.(20 volumes of 200gb).

you need dedicated HBA for your DISKPOOL(We had 4 HBA port dedicated for the diskpool in a redundent SAN infrastructure) The rest were for the tapes..
Do not mix TAPES and DISK on the same HBA. (NOT RECOMMENDED)

 

[spiffy]

My Db and Log volumes reside on SAN storage also, i could probably use that HBA also i am thinking, as it will all be going to the Symmetrix? I think the disks will be spanned over 3 or 4 luns, but am not 100% on that.

 

[psefmoreit]

Spindles

How many raid groups does your 4TB span across in the Symmetrix? You would want to create at least one diskpool LUN on every raid group. This would spread the load between each group of disks (raid group).

 

[n9hmg]

I recommend you make your DISK-class pool only large enough to hold a days-worth of sub-GB files (with a 30% cushion), and use the rest of the disk to make a FILE-class pool set as NEXT for the DISK one, and set maxcap on the DISK pool to shunt the larger files straight to FILE and let the random-access DISK pool do what it does best - lots of little stuff. That way you can clear your DISK pool daily, for better performance.

 

[tenpin]

I have 40TB of ds4800 SAN disk at one site and about 60TB at another, each servicing one TSM server. My primary storage pools are devclass=file so I run small storage pool volumes, 128GB each though in a smaller environment such as yours I would probably use devclass=disk since devclass disk allows simultaneous writes to stgpool volumes.
I would carve up your disk into 20GB volumes, this will allow you to move volumes to diff. storage pools if the need arises.

My servers run 4 dedicated HBAs each for storage pools and the disk itself is carved up into vertical LUNs on the ds4800. I would suggest running your storage on a dedicated HBA, mixing disk and tape on the same HBA will almost surely cause you some grief.

If possible have your database and log volumes on separate storage, or at least separate LUNs and run them on their own dedicated HBAs...

Check the implementation guide for more ideas on storage pools. And remember adapter placement in your server can have a big impact on performance.

 

[spiffy]

I just plan on keeping the 1 diskpool i have set up on each TSM server, add the new disk, then just define x number of 200GB volumes.
This is more of a bandage for now, as further down the road, there is talk of redesigning the TSM structure.
This is more to just keep 24hrs worth of backups in one place before tossing it to tape.

thanks for all your hints and tips

 

[spiffy]

I think i will take the advice to move large client back ups(sql and exchange) directly to tape. i only have about 5 very large ones that can go to tape directly..

thanks
James

 

[batu]

All,

Did anyone upgrade their system to TSM 6.1? What is the tape library compatibility requirement for TSM 6.1? We have two Spectra Logic libraries with Blue Scale 10.4.0 and 10.0.7.