Video Screencast Help

Large MSDP pools VS 3-4 smaller pools

Created: 21 Feb 2013 • Updated: 27 Mar 2013 | 1 comment
This issue has been solved. See solution.

I'm trying to Gauge the pro's and cons of the following setup configs:

Config 1

2 Media servers each with thier own 32 TB MSDP pool (64 TB total)

Config 2

4 Media servers each with 16 TB MSDP pool (64 TB total)

basically the question boils down to what are the pros and cons of spreading available storage across multiple MSDP pools.

Operating Systems:
Discussion Filed Under:

Comments 1 CommentJump to latest comment

RLeon's picture

Notes on deduplication scope:
Deduplication happens between data under the same node (MSDP Storage Server), and not across multiple nodes.
So the more nodes there are, the more overall storage you will end up using because data already present on a node will need to be stored again in other nodes - If you duplicate or inline-copy all your backups to multiple nodes for HA.

When NetBackup documents talk about "Global dedup", it is really referring the scope within a single Storage Server.
It can get confusing when people say that "Global dedup works across nodes in the 5000/5020 appliance".
The 50XX series can be stacked together to increase the total storage space, but it still presents itself to NetBackup as a single Storage Server, therefore the single node/Storage Server dedup scope still applies.

Less MSDP nodes but more storage on each:
Less individual dedup scopes means less overall storage consumption - If you keep copies of the same backups on multiple nodes. If not then this point is not relevant.
Environments with less servers ease management.
Node specs will need to be more beefy (Processor, RAM, SAN/LAN bandwidth, etc), to handle the larger dedup storage and heavier backup/restore responsibilities.
Less nodes to store extra backup image copies means less protection.
If the stoarge come from the same SAN array, then you could be stressing the same array, or worst, the same disk raid group(s). No storage performance benefit here. Also, less protection.

More MSDP nodes but less storage on each:
More nodes to store extra backup images means more protection to your backups - If creating multiple backup copies between nodes etc.
Backup and media-server-deduplication loads could be shared/balanced between nodes for better performance. (Better node and network/SAN specs mitigate this...)
Less specs requirments on individual nodes.
If the storage comes from different SAN arrays/sources, then more storage performance (potentially) and no single point of array failure; better HA.

More servers to manage.
More individual dedup scopes means more overall storage consumption - If using multiple backup copies etc.

Personally I tend to recommend less nodes but with high specs on each (including better LAN/SAN infrastructures), unless someone really wants many copies of backup images to be stored on many different nodes for that extra protection.
But then again, regardless of config1 or config2, some people insist on not making "redundant" extra backup copies between nodes. I stay away from these people.