on-host backups

[blodsbror]

Hi,

We are changing storage providers soon (Compellent), and in the meantime need to run our backups on-host. I'm just wondering how much influence the CPU/Memory on each host, has towards prcessing backups ?. Or is the speed of a backup (among other things such as compression etc) mostly relying on the speed at which each host can read from the SAN/CSV ? Any tips to increase on-host backup speed, would be appreciated!

At the moment, I'm using the MS VSS provider, Optimal compression and LAN target (as backups go onto local disks on another server).

cheers!

[veremin]

In terms of consuming system resources the compression activity is the most intensive one. However, with Optimal level it shouldn't bother you that much. Please, refer to the corresponding section of our Help Center for other system requirements.

As to the speed increase, only the actual tests with full bottleneck statistics should give the final answer on how the current performance can be improved.

Thanks.

[Gostev]

Hi

I would like to warn you that we've once seen on-host backup mode cause issues on over-provisioned Hyper-V clusters (think big public cloud provider environment). The issues appeared immediately after temporary switch from off-host backup mode (that was deployed by our Solutions Architect) to on-host backup mode to workaround some minor issue, and this caused significant cluster stability issues.

As such, I would not recommend on-host backup mode for hosts and clusters running at near full capacity. However, it is perfectly fine for environments with some spare resources (which is what most environments are like), and this backup mode is what majority of our users have deployed.

Thanks!

[blodsbror]

Hi!

Thanks for the replies. I'm forced to use on-host as we were using a trial license for HP snapshots on our EVA SAN, which has expired. And this won't be renewed as we are receiving a Dell Compellent in approx 3 weeks. So off-host with a hardware provider isn't possible until we move our data over to the new SAN (which is licensed for snapsots).

There isn't a lot I can do here perhaps, apart from sending in the bottleneck logs perhaps. CPU on each host isn't going past 65% or so, and stability has seemed fine so far.

[veremin]

If the production host gets overwhelmed with backup/replication tasks, you can always decrease the workload, limiting the number of concurrent tasks the on-host proxy or backup repository can handle. Thanks.