Performance Differences between 2 nearly identical bkp jobs

[bunger]

Here is the environment:
3 Dell PowerEdge R710 servers, all identically configured with 24GB RAM, dual E5530 Xeon procs, and 3TB of local storage
1 Dell PowerVault MD3000i iSCSI SAN
2 Dell PowerConnect 5424 switches

Each ESX box is called ESX1, ESX2, ESX3
All production vms reside on the SAN with processing handled on individual ESX boxes
vCenter is virtualized on a Win2k8 server

ESX1 manages vCenter
ESX2 manages FS1 AND FS2

Veeam is running inside vCenter on ESX1

Veeam has 2 backup jobs:
backs up FS1 from SAN to local storage on ESX1
FS1 size is about 950GB
backs up FS2 from SAN to local storage on ESX3
FS2 size is about 230GB
both jobs are configured identically, aside from the target storage being ESX1 and ESX3

FS1 runs Win2k3, AD, and Exchange
FS2 runs Win2k8 and SQLServer

The backup for FS1 takes 11-14 hours with an average processing rate of about 20 MB/s
The backup for FS2 takes about 8 minutes ( after the initial backup ) with an average processing rate of about 470 MB/s

I have looked everywhere and cannot figure out why there is such a large disparity between the two jobs....

Thanks in advance for any help!
BIll

[Vitaliy S.]

Hello Bill,

Could you please verify that CBT is enabled for the first backup job? If it is enabled, then please check if it does work for VMs being backed up?

You can easily do that by looking at real-time statistics while virtual disk (VMDK) is processed. Another way would be to check it with our support team via investigating corresponding job log files.

Thank you.

[bunger]

CBT is enabled.

I restarted the backup job and watched the stats, which report: Backup mode: HOTADD/NDB with changed block tracking

[J1mbo]

CBT needs the VM to be on V7 virtual hardware, btw. Also, Exchange could be the key factor here, as it happens I just posted about that. It runs defrag tasks by default overnight for four hours which touch (change) a huge number of blocks. See http://blogs.technet.com/b/exchange/arc ... 04504.aspx.

[bunger]

VM versions are 7 for both.

Regarding Exchange, it is exchange 2003....

[Vitaliy S.]

I restarted the backup job and watched the stats, which report: Backup mode: HOTADD/NDB with changed block tracking

Try choosing local storage on ESX3 to see if makes any difference or not. Besides, this job might also failed over to NBD processing mode.

Also, Exchange could be the key factor here, as it happens I just posted about that.

That's true. Please expect to have different processing rates for different VMs, see details in this post: Very different backup speed on VM's

Regarding Exchange, it is exchange 2003....

Well... that explains, here is an existing topic for more info: Slow VM backup with SQL Server 2008 R2 (nevermind the title)