I'm hoping someone will read this post, take pity on me, and give me some good advice.
I'm filling in for our Veeam admin while he is on long term sick - possibly not having been helped by the struggles he's had with our Veeam installation. So I'm on a steep learning curve.
We are running Veeam 8, and have licensing for 15 VM hosts (although we have more). Our hosts run vSphere 5.1
We have 2 virtual Veeam proxies, a physical Veeam repository (8 cores, 32GB RAM), and an offsite physical repository (currently broken) for replication. A tape robot (currently broken) is connected to the primary Veeam repository.
It is also connected to a mixture of 5 lumps of storage - iSCSI and fibre channel from 4 SANs (although one of the two FC HBA's in the server is broken) . I'm attempting to add a 6th volume, but the SAN isn't yet connected to the SAN switch. The storage volumes range from 4TB to 14TB, and we're using Windows 2012 deduplication to try and squeeze everything in, which is proving a challenge.
I've noticed the network card on the primary repository is maxed out during backups, and the Windows 2012 deduplication is unable to keep up with the amount of backups being done each week.
We're running daily incremental backups (no synthetic fulls), with active fulls created the first Saturday of each month (although I notice some jobs are doing fulls every Saturday, and there is the odd job with weekly Synthetics). Monthly copies are supposed to go to tape (but not since August!!).
I'm thinking that upgrading to Windows 2012R2 will give a performance boost for the deduplication, and Veeam 9 will help with managing all the volumes. I was also planning to stop using Windows Deduplication on the fastest volumes, so we can spin up VMs from the backup storage in the event of a problem (which was why we bought Veeam), with copies to the deduplicated storage for long term retention.
I was also wondering whether it would be a good idea to create 6 virtual repository servers - each with one of the volumes to aid the deduplication processing, and remove some of the network bottleneck, leaving the current physical server as just a remote tape server (when I get the tape fixed!!).
Any advice would be greatly appreciated.