PowerNFS Speed? Can i improve it?

[robertorcp]

Just curious if i was doing something incorrect or just have to deal with the speed of it.
Physical Veeam Backup Server has ISCSI shares were the backups are stored. If i do a surebackup and try to bring up a server with 700GB its very slow.
What is the optimal configuration of PowerNFS. Should i use physical local disks instead of Iscsi?

[foggy]

Roberto, what kind of storage device do you use for storing backups?

[robertorcp]

I am using ReadyNas 4300 - 12 SATA drives (7200rpm)

[foggy]

Am I understanding right that it is not a dedupe storage, is it?

The following topics might be helpful in your case: SureBackup Powering on Taking Forever and Very slow application testing in Virtual Lab

Btw, what hardware are you using to run vPower NFS datastore?

In any case, debug log files could point out the bottleneck.

[robertorcp]

So here is the exact configuration:
I have a DL380 G6 with 2 nics cards (one for iscsi, one for Lan) (this is the veeam backup server). The iscsi connects to the readynas for the backups. The veeam job for Virtual Lab connects to the PowerNFS on the LAN side because we have a vkernel for the LAN. (1Gbs) Therefore the path it takes to show the storage is first connects to backups over the iscsi port then vmware connects to the powernfs share on the lan port. Are you getting the picture?
Whats a good search key to use for the log files, they tend to be very long!

[tsightler]

PowerNFS performance is generally primarily limited by the random read I/O performance of your storage device, and secondarily by the CPU performance of your PowerNFS host. The ReadNAS devices aren't known to be very fast, especially for random I/O. You might want to try running a simple benchmark like CrystalDiskMark and seeing what performance you get from the "512K" test. This will be roughly what you can expect from PowerNFS, minus a little bit for the overhead of our compress/dedupe.

[robertorcp]

I will do that this afternoon and post, thanks!

[robertorcp]

I did the Crystal Disk Mark on the ISCSI drive attached to the server and the results were:
512k Random Read 1GB dataset - Read 84.98 MB/s | Write 90.14 MB/s
I think those are decent numbers?

[tsightler]

Can you run a larger dataset as I'm surprised at those numbers being so high. I'm wondering if you are seeing a cache effect.

Also, do you have the latest V6 patch installed?

BTW, you also aren't very specific regarding your issue. When you say "it is very slow" can you clarify? Are you performing an actual restore (with network enabled) or are you comparing restoring a disconnected VM to the normal boot time?

[Fizzer]

[merged]

Hi All,
am playing with my new Veeam setup and have succesfully backed up/restored a few bits and bobs.
I'd like to make a really fast surebackup environment to bring back multiple VM's to extract sharepoint data.

we have 7 esxi 4.1 servers - 4 production and 3 test/dev
we have 2 netapp filers 1 local to prod and 1 local to test/dev
test/dev and prod have Gb connectivity
repository is an netapp iscsi disk presented to VM - (backed by 15K sas drives)
redo disk is an NFS datastore presented to host (backed by 7.2K sata)

I am thinking of nominating a single esxi host in the test/dev to be the primary target for bringing stuff back (surebackup, uair etc.) I have tested that this works OK

I was considering installing a pcie ssd card in this host to run the redo storage.
I also think I should have a proxy on that machine

would that be a good idea? Anything else that I could do to speed it up?

Thanks
Andy

[foggy]

Andy, besides the above discussion, maybe this one will also be useful to you.

[Vitaliy S.]

Deploying additional proxy server will not improve SB job performance, but storing redo logs on SSD disks will definitely help. Thanks!

[Fizzer]

Checking online, SSD support for perc6/i controllers and vmware is very limited at the moment (without lots of cash) so I've just installed 4 x 15K 146 GB sas drives I had knocking around into an esxi host in a Raid 0 set.
I'm re-running my surebackup job now using this local raid 0 set for redo logs, the first DC power on time has reduced from 12:14 to 8:30
That should make a reasonable difference overall. I'd like to do better

At the moment I have 3 DC's, an SQL server and a 4 server sharepoint farm in the same surebackup job. This takes over an hour to run so any savings I can make are a bonus. I'm planning to reconfigure the job to boot only the DC's then try to use powershell to boot the SQL and SP servers into the correct network.

Anyone else got tips to improve performance?