We are the proud owner of 4 NBU appliances (2 x 5230 on PROD site and 2 x 5330 on DR/DEV site, these 2 sites are running OptDup from PROD to DR).
We recently bumped up the storage on the 5330 appliances and I'd like to configure them to utilize Accelerator and get the best dedup rates possible. We have an SUG with the 2 x 5330, set with Failover - the problem we had would be the appliance in the top order would fill u,p and then backups would happen on the second appliance. This caused some servers that rely heavily on Accelerator to work its magic (eg on slow / small WAN bandwidth) to perform backup on an appliance (the second one in this case) where the track log doesnt exist/not updated and backup would run really, really slow then.....I hope that makes sense.
I'm hoping to tackle this issue by either :
1) Creating multiple copies of the backup via SLP - 2 copies to be exact, each going to an 5330. My concern with this is that I found a technote that says (to my understanding), that this might not work :
2) Creating one copy of backup via SLP to one 5330, and afterwards create a duplicate of that copy to the other 5330. My concern with this is that I'm not sure whether Accelerator/track log would be created on that media server doing the duplicate.
Just wanting to know if anyone has ever attempted to do something similar, and if so, how? My goal is to ultimately have an Accelerated SUG containing 2 x 5330 that will both be able to be used by an Accelerated backup client interchangeably without performance impact even if last full backup was done by one and the next one done by the other.
My master server is a Windows 2008 R2, and all the media servers are the appliances (2x5330 and 2x5230) as mentioned above, all having NBU 7.7.3
Hi Sue - AFAIK the track logs for plain client backup are not stored on media server, I always thought that they are on the client in form of:
...so, yes, you are right... multiple track logs for different MSDP based storage units because MSDP are islands and a storage server can have only one MSDP pool.
So, the two behaviours go hand-in-hand so to speak... with your question of track logs, and the linked post of dedupe-ness... so, as I see it, it is still an "affinity" question... and so for me, it's a useful reminder and another reason for us admins to carefully consider not using multiple storage units in storage unit groups for Accelerator based backups. I mean still use storage unit groups, but with only one storage unit member, so that we can control to some degree "affinity" between storage-server and backup client. I don't like doing it, but it does mean more predictable backup job behaviour.
And if we stop and think about how dedupe works and how accelerator works then there really is nothing that can be done by any software vendor... because the advanced techniques of thinning out data are really actually tied very tightly to the storage side... and so clients must track the optimizations that they have previously been able to negotiate with platforms of backup storage. So, this isn't so much a limitation, it is actually a feature of all backup products and all backup storage wherever there is a delineation of storage server.