HotAdd and Performance

[hwaccend]

Hi all

I have one host(Dell PowerEdge M710HD,8*CPU-Intel E5620.2.4Ghz,Vmware ESXi 5.1.1 ),two VMs(2vCPU with 4vCores, 16GB RAM,windows 2008 Enterprise R2 x64) on the host .I just installed Veeam B&R(V6.5) on one VM and I wanted to backup another VM, the two Vms share the datastore from SAN storage through 10GE iSCSI.

I also intalled Veeam B&R on anoter physical sever as proxy(Transport Modes: Direct SAN Access), The porformce of SAN is about 150MB/s, the Network mode is about 50MB/s ,but the storage is very free(about 40%-50% usage of storage disks) .Everytime I run the backup, bottlneck is always: source.

Questions:


1,How can I improve the backup performacne when SAN and Network mode?

2,When I configure the transport mode as Virtual Appliance, but it allways switch to nbd(not hotadd), How to configure this ? Please give the detail topology?


Bellows are the information from the backup job:

4/16/2013 5:20:48 PM :: Queued for processing at 4/16/2013 5:20:48 PM
4/16/2013 5:20:48 PM :: Required backup infrastructure resources have been assigned
4/16/2013 5:20:50 PM :: VM processing started at 4/16/2013 5:20:50 PM
4/16/2013 5:20:50 PM :: VM size: 70.0 GB (34.7 GB used)
4/16/2013 5:20:52 PM :: Using source proxy VMware Backup Proxy [nbd]
4/16/2013 5:20:59 PM :: Creating VM snapshot
4/16/2013 5:21:12 PM :: Saving '[win2008-0S] windows2008-02/windows2008-03.vmx'
4/16/2013 5:21:12 PM :: Saving '[win2008-0S] windows2008-02/windows2008-03.vmxf'
4/16/2013 5:21:13 PM :: Saving '[win2008-0S] windows2008-02/windows2008-03.nvram'
4/16/2013 5:21:14 PM :: Hard Disk 1 (50.0 GB)
4/16/2013 5:27:30 PM :: Removing VM snapshot
4/16/2013 5:27:40 PM :: Swap file blocks skipped: 16.0 GB
4/16/2013 5:27:40 PM :: Finalizing
4/16/2013 5:27:46 PM :: Network traffic verification detected no corrupted blocks
4/16/2013 5:27:46 PM :: Busy: Source 99% > Proxy 20% > Network 1% > Target 4%
4/16/2013 5:27:46 PM :: Primary bottleneck: Source
4/16/2013 5:27:46 PM :: Processing finished at 4/16/2013 5:27:46 PM



Thanks for help

[Vitaliy S.]

Hi Victor,

1,How can I improve the backup performacne when SAN and Network mode?

You say that direct SAN mode is giving you 150 MB/s which is not bad at all Is it for the full run or incremental job pass? Anyway, we've got a couple of tips on how to improve direct SAN mode (iSCSI) job performance, please check them out.

As to the network mode, then since all traffic goes through ESX(i) network interface, there are not many things you can do to improve that. I would say that 50 MB/s is somewhat expected for jobs running in network mode.

2,When I configure the transport mode as Virtual Appliance, but it allways switch to nbd(not hotadd), How to configure this ? Please give the detail topology?

Where is your VMware Backup Proxy located at? Are you running it on the ESX(i) host which has access to all datastores where you backup virtual machines from?

Please check our sticky F.A.Q. for the ideas on why HotAdd mode might failover to network mode.

Thanks!

[hwaccend]

Hi vitaliy,

The VMware Backup Proxy is located at one VM. I want to backup another VM by HotAdd mode. But how to let the proxy VM to access to all the datastores? should I add any virtul disks to the Proxy?

[foggy]

If both VMs reside on the same host, no additional configuration is required. You can just try and run a backup job using that proxy and see whether it uses hotadd to backup the second VM (the corresponding [hotadd] tag will be displayed right after the proxy VM name in the job statistics window, after selecting the VM being backed up in the left pane).

If this does not occur, please refer to the following KB article: http://www.veeam.com/KB1054 containing several limitations of using hotadd mode, probably the issue could be in one of them.

[hwaccend]

Hi foggy,

I re-checked the configuration refer to the KB article. The restored VM results in the problem, I can use the HotAdd mode,now.
But everytime ,the Primary bottleneck is Source (99%), according to my configuration(refer to above), which component is really the bottleneck?

[foggy]

Victor, the bottleneck stats show the percent that every component involved in the backup process is busy so there will always be a bottleneck in every setup. Bottleneck source indicates that the disk reader component spends all of the time reading the data, because the following stages are always ready to accept more data for processing. Please refer to the sticky FAQ topic for further explanation.