Backup Speeds

[homerjnick]

Can I ask a quick question to see if I have this right?

We have setup:

Netapp N Series 3600 SAN with 16TB of FC and SATA disks on RAID-DP (equivalent to RAID 10).
IBM BladeCenter with 4 Blades...3 ESXi5 servers and 1 physical W2008 box that is our Veeam server v6 latest patch.

I have about 16 LUN's configured as Datastores in our ESXi environment and the physical Veeam server, in the same Blade Center, can see these LUN's in its Disk Management but obviously not initialised (shudder).

The Veeam server has a 3TB LUN assigned to it that is on the same Production SAN.

Therefore, I assume Veeam has direct access via storage to the Datastore LUN's and that backups would be very quick but I only get 11MB/s when backing up VM's.

I get 22MB/s when backing up VM's to a NAS box...

How can I diagnose what the issue is? Surely the backups from the physical Veeam server, which can see all ESXi LUN's to its backup LUN, which are on the same SAN, should not be slower than backing up to a network device???

[foggy]

Nick, the first thing to look at is the bottleneck stats numbers provided for each job in the Realtime Statistics window (available on the job's shortcut menu). Also, you can check there which transport mode is effectively being used by your jobs during backup (just select the particular VM to the left in the same window to get this information).

[homerjnick]

A backup onto the SAN bottleneck stats shows:

Source: 55%
Proxy: 21%
Network: 25%
Target: 83%
Processing Rate: 11MB/s

A backup to the NAS shows:

Source: 43%
Proxy: 73%
Network: 27%
Target: 77%
Processing Rate: 24MB/s

Where does it tell you about the transport mode? For a VM in my job window I have:

10/05/2012 01:38:47 :: Using source proxy VMware Backup Proxy [san;nbd]
10/05/2012 01:39:17 :: Hard Disk 1 (15.0 GB) 484.0 MB read at 3 MB/s [CBT]

[foggy]

Source: 55%
Proxy: 21%
Network: 25%
Target: 83%
Processing Rate: 11MB/s

According to this, your target is the bottleneck. This either could be explained by the storage performance in general or the fact that it is used as both the primary storage for your VMs and backup repository, which applies additional load on the disks.

Btw, what backup mode do you use in your jobs? Note that reversed incremental mode puts 3x I/O load on target comparing to forward incremental.

[homerjnick]

Reversed Incrementals it is...these VM's are also on second tier storage ie SATA so perhaps performance is as good as it will get for these VM's...

[dellock6]

Also remember you are using the same storage both for running VMs and for Veeam, so you creates much more I/O on the NetApp when you run your backups, the load of extracting vmdks from the production LUNs, and the 3 I/O of the reverse incremental.
At least you can check if NetApp is using the same areas for production and backup luns, and try to separate them in order to use different spindles...