Container Pool size recommendation

[ch_wira]

Hi,
I have a plan to upgrade TSM server to be 8.1 and using container pool instead of VTL & Disk base. Any idea if I plan create only one container pool (1 PB raw capacity)? My objective is high deduplication ratio expected.

 

[Trident]

Hi,

https://www.ibm.com/developerworks/... Manager/page/IBM Spectrum Protect Blueprints

The largest system IBM blueprint has 1 Pb storage.

You should consider creating two instances.

 

[ch_wira]

Hi,
Thanks you. By the way what's about the big one size of container pool. Anyone can share pros and cons. I found only one factor is the protect pool will perform a whole pool cannot select specific data. Is there any performance impact?

 

[Trident]

The size of each stgpooldirectory is something to consider. How many luns/paths can you handle?
What happens if you loose one or more luns (for whatever reason)? The time to recover it can be interessting. And the db2 queries to put extenst to damaged/destroyed is a big operation.
A big dedup pool will demand a lot from your DB2 part. Can your system handle it? The layout of db luns and log luns is becoming more impoortant.

Just some thoughts from my side.

 

[marclant]

The size of each stgpooldirectory is something to consider. How many luns/paths can you handle?
What happens if you loose one or more luns (for whatever reason)? The time to recover it can be interessting. And the db2 queries to put extenst to damaged/destroyed is a big operation.
A big dedup pool will demand a lot from your DB2 part. Can your system handle it? The layout of db luns and log luns is becoming more impoortant.

Triden brought up a lot of good points.

I recommend you stick to the BluePrint: https://www.ibm.com/developerworks/... Manager/page/IBM Spectrum Protect Blueprints

As well at Config Best practices from the manual: http://www.ibm.com/support/knowledgecenter/SSEQVQ_8.1.0/perf/t_optim_config.html

 

[ch_wira]

I have reviewed the BluePrint & Config Best practices as marclant's recommendation but not found any related to storage pool size.

As Triden points.
- How many luns/paths can you handle?
If this related to the number of folder/files filesystem limitation. I agree to split them to more filesystems but what's about big one storage pool because we can add more LUNs (filesystems) to a single storage pool.

- A big dedup pool will demand a lot from your DB2 part.
Can you share that why DB2 require more resource if the storage pool is big? Is it related to deduplication table in DB2 that contains the meta data?

 

[marclant]

I have reviewed the BluePrint & Config Best practices as marclant's recommendation but not found any related to storage pool size.

It sounds like you are aiming for the large configuration, here's from page 11 of the blueprint:

https://www.ibm.com/developerworks/...-ccd2b5cac354/media/srv_blueprint_aix_v22.pdf