Improving direct-from-SAN backup speed with iSCSI SAN

[jbush]

Alright, now that I have the MPIO issue resolved on the production SAN, Dell Equalogic PS4100. Issue was with the Dell Multpath Extension Module.

Anyways, here are the more conclusive results of having autotuninglevel set to normal versus disabled on a VM running Veeam with direct SAN access.

autotuninglevel=normal
Total VM size: 760GB
Processed: 496.2GB
Read: 491.4GB
Transferred: 343.3GB (1.4x)
Total duration: 1h19mins
Processing rate: 115MB/s
Busy: Source: 91% > Proxy 41% > Network 13% > Target 11%

autotuninglevel=disable
Total VM size: 760GB
Processed: 496.2GB
Read: 491.4GB
Transferred: 343.4GB (1.4x)
Total duration: 1h14mins
Processing rate: 123MB/s
Busy: Source: 94% > Proxy 43% > Network 10% > Target 11%

Source server details:
Windows Server 2012 R2
Exchange Server 2013
Disks, 1x60GB, 1x400GB, 1x300GB
vCPU: 6
vRAM: 32GB

Veeam server details:
Server 2008 R2 SP1
Veeam B&R v7.0.0.715
vCPU: 12
vRAM: 32GB
Using MS MPIO for Production iSCSI Fabric, isolated Force10 S25 switches for iSCSI traffic only
Using MS MPIO for Backup iSCSI Fabric, isolated vlan, HP 2920 switch
MTU 9000 across the two fabrics

Production SAN details:
Equallogic PS4100
RAID: RAID50, 3 strips
HDD: 24x 900GB SAS
Controllers: 2x1gbps iSCSI (one controller standby, one active, 2 NICs per controller)
Hardware iSCSI initiator

Backup NAS details:
QNAP 1279U 12bay
RAID: RAID5, 3 strips
HDD: 12x 3TB SATA
Controller: 2x1gbps
Software iSCSI initiator

[dellock6]

First of all Jason thanks for posting the results!
Seems like indeed disabling autotuning brings an advantage, but it's a small one. Something to take into account when deciding to disable it or not (better performance vs. default configurations).

Thanks,
Luca.

[SilentDeath]

Hello All,

I'm using Veeam DIRECT SAN access aswell.
The VMware VMDK's are stored on HP Lefthand SAN LUN's.
My back-up jobs run approximately at 55 MB/s.
Last night I modified the autotunig level but this didn't make any difference.
Anyone any idea how I could speed up jobs? I'm storing back-ups locally on back-up server.

Let me copy the template from jbush:
(normal incremental job, active full in weekend)

Processed: 1.2 TB
Read: 49.1GB
Transferred: 18.8 GB (2.6x)
Total duration: 18:55
Processing rate: 55 MB/s
Busy: Source: 99% > Proxy 51% > Network 3% > Target 0%

Veeam server details:
Server 2012 R2 SP1
Veeam B&R v9
CPU: E5-2620 v3 @ 2.40 GHz (6CPUs)
RAM: 32GB
HP Proliant DL380 G9

Production SAN details:
HP LeftHand SAN

Backup NAS details:
Data is stored locally on backup server.

[joergr]

Hi,

looking at your data neither the network nor the target is the culprit / bottleneck. Seems to me that your SAN just simply can´t deliver more. See for yourself: Source is at 99%.

Thanks

Joerg

[foggy]

What kind of Lefthand device do you have? Wondering if you can use storage snapshots.

[PTide]

Could you please double check if your job utilizes Direct SAN mode and not failing to network mode? You should see something like this

"6/6/2016 2:02:51 PM :: Using backup proxy WIN-LD86G2UPHMK for disk Hard disk 1 [san]" in session log.

Thanks