Infrastructure upgrade recommendations

[Ross1976]

Hi all,
we need to revamp our current Veeam infrastructure (up and running since late 2012) because our business workload has dramatically changed and the amount of daily incrementals is seriously going up - and backup performance is slowing down. So I'm asking you some recommendations on how to improve speed and reliability.

We presently use Veeam B&R 9 to backup 50 VMs (Vmware ESXi 5.5) to a FAS2240 SATA backup repository (local iSCSI volumes mapped through a Windows 2008 R2 Storage server). We have a W2008 R2 hot-add proxy. Production storage is another beefy FAS2240 (FC storage) with a mix of SAS and SATA aggregates. We also replicate all of the VMs to a remote DR site (100 Mbps link). Local backups have been running fairly well but we now struggle with bigger fileservers because of their large incrementals (always in the 100-150 GBs every weekday) and finalization may take several hours. This of course impacts replication and jobs can even overlap.

So, we were wondering how we could improve backup speed, finalization and decent replication. We were thinking about storage-array based snapshots, or just simple DirectSAN backups. Maybe we could add another beefy proxy and backup repository. For decent replication times, we could upgrade the current link to the DR site.

Thank you for your valuable suggestions.

[foggy]

All your thoughts are valid and could allow to improve backup performance. However, where you should start analyzing your setup to find ways of optimization is bottleneck statistics reported by your jobs. Have you reviewed them?

[Ross1976]

Foggy,
in most backup jobs the bottleneck is Source (sometimes 80-84%, sometimes 90%). In a couple of jobs, the bottleneck is Target (70%).

[foggy]

For those reporting source as the bottleneck, using storage snapshot (or trying direct SAN) is the way to go. To address target being the bottleneck, you should pay attention at the repository (switching to forward incremental will help, if currently reversed incremental is used).

[Ross1976]

Thanks.
We do use forward incremental for all of our jobs. The NetApp integration looks promising and hopefully backups can be faster. So we need a new physical proxy with a FC HBA that can access the production datastores.

[foggy]

Backups will be faster due to faster VM snapshot commits, and production VMs will be less affected as well.

[marcusa]

San based Array snapshot is the best way to go for speed if you have TB's of data