Disk pool deduplication other than TSM container pool

Mita201

ADSM.ORG Senior Member
Joined
Apr 20, 2006
Messages
601
Reaction score
31
Points
0
Location
Beograd, Serbia
PREDATAR Control23

Hi,
does anyone has some experience with using other means of reducing data in TSM storage pools than TSM container pools? Like doing it with backend storage dedup/compression capability, or even with some file system utilities (latest redhat maybe, or other)?
 
PREDATAR Control23

Hi,
does anyone has some experience with using other means of reducing data in TSM storage pools than TSM container pools? Like doing it with backend storage dedup/compression capability, or even with some file system utilities (latest redhat maybe, or other)?
You mean like the legacy file pools with deduplication and identify duplicates running? Or are you referencing storage arrays that can compress and dedup like Pure or other vendors do?

Sorry, just asking for clarification :)
 
PREDATAR Control23

I think he is talking about some sort of "Storage Dedup and/or compression" something like VTLs do.
 
PREDATAR Control23

I am talking about "external" features, of storage, or maybe file system, not ISP/TSM functionalities.
The thing is, as container pools introduced some great space savings and maybe some benefits in replication process, they also removed great flexibility that we have with "traditional" pools, and also made ISP server unreasonably hungry in terms of needed CPU/Memory and database IO resources.
When I am talking about flexibility, you were able to migrate data from one pool to another, to copy them in an efficient way, to do backupsets and so on.

About resources: I am supporting, as a consultant, several ISP servers here and there, mostly of "small size" according to IBM ISP blueprints, some of them maybe in "middle" segment. Recently I have upgraded couple of them to latest server versions, migrated data to container based pools, set replication and even container copy to tape in few places, one on DR side, and another (client) on primary side.

As backup operations mostly goes quite ok, I can see enormous growth in hardware usage: Database grows 3-4 times, ISP servers are the hungriest IO consumers of ALL servers in (small) data centers. OLTP and OLAP systems of a bank are no match to ISP database regarding storage IO, which is crazy.

Replication (container pool then node replication) is somewhat buggy, but this is a different story

So, I would like to explore other options on data reduction, beside ISP functionality.
 
PREDATAR Control23

Hi,
If you precompress data client side, whatever storage is not going to do any data reduction. We have about 25 - 35% reduction due to compression. Nice reduction on license cost.

If you encrypt data client side, whatever storage is not going to do any data reduction whatsoever. Compress, then encrypt clientside works nicely. Watch cpu usage .......

Sending data without compression will add lots of data to your license cost, bandwidth usage goes up.

Finding a storage that will compress/dedup is not going to be cheap either.


Have a look at your license model, your data being backed up and a few possible storage vendors. Some excel charts, and you have a solution.

I know, not a solution, but some key points to look at.

-= Trident =-
 
PREDATAR Control23

Hi,
If you precompress data client side, whatever storage is not going to do any data reduction. We have about 25 - 35% reduction due to compression. Nice reduction on license cost.

If you encrypt data client side, whatever storage is not going to do any data reduction whatsoever. Compress, then encrypt clientside works nicely. Watch cpu usage .......

Sending data without compression will add lots of data to your license cost, bandwidth usage goes up.

Finding a storage that will compress/dedup is not going to be cheap either.


Have a look at your license model, your data being backed up and a few possible storage vendors. Some excel charts, and you have a solution.

I know, not a solution, but some key points to look at.

-= Trident =-
Hi,

Thank you for your answer.

Yes, I am aware of all the problems it brings, with license (if front end model is used), and other, as well. I am just checking if someone has real world experience, like yours, with client side compression.

Regards,

Mita
 
PREDATAR Control23

We have experience with a vendor that offers deduplication/compression in a storage appliance. The vendor actually shipped us free storage because their audit of our environment said they will guarantee a specific dedup/compression ratio if we went with them. So we did, and less than stellar results were achieved. This storage system isn't presented to our TSM server in any way, other than I back it up.

Other than that, I let TSM handle all the dedup/compression (client side mostly). There's something warm and fuzzy about knowing the duplicate information is backed up in a database that I control, vs some metadata file on the storage appliance that no one ever sees other than the appliance vendor. Yes, I'm very aware of some problems in the past regarding deduplication, and have been burnt by them but over all...still pretty happy with it.

Be wary of a storage vendors stating you will achieve X ratio with hardware. Get it in writing, and get some sort of clause that they will provide additional storage if their system cannot provide the reduction stated (down side is you are now paying maintenance on extra shelves).
 
Top