Pinpointing Bottlenecks

[PeakSoftware]

Does anyone have any suggestions for pinpointing the bottleneck in backup processes? The report shows my source is the bottleneck, but I'm not quite sure where to go from there.

[foggy]

George, that means that source storage data retrieval speed is a bottleneck. What is your setup and what transport mode do you use to read the VMs data from the storage?

[PeakSoftware]

I am running Hyper-V and the VHDs are on an iSCSI storage device directly attached to the host server. I have done some HDD throughput tests and the results are nearly the same as local storage.

[PeakSoftware]

BTW, these are incremental backups as where the first full backup shows the network to be the bottleneck at 98% which is what I would expect going over a VPN to an off-site location. Is it normal to see the source as the bottleneck at 99% with an incremental backup?

[Gostev]

Yes, especially if CBT is not used.

[PeakSoftware]

CBT is enabled.

[Gostev]

"Enabled" is not equal to "Used" though
Do you see [CBT] letter next to each virtual disk in the real-time statistics?

[PeakSoftware]

Good to know. Yes it is there. It also shows it was read at 4MB/s. Is there a place I might be limiting the job to that 4MB/s? My repository is limited to 3MB/s because it's going over a WAN, but I don't see anywhere else to limit a data ingestion/througput rate.

[Gostev]

Bad random I/O performance of the storage, for example.

[mongie]

I've been trying to pinpoint the bottleneck in my system - i have a similar issue.

The random read/write you speak of - what block size would be used?

[Gostev]

No idea... but you can launch Process Monitor and see for yourself.

[dellock6]

If your storage has some kind of management console, you can take a look at it and see the performances during backup activities, like IOPS or MBs. Also, knowing model and disks could help to understand the estimated speed and check it against the real speed.

Luca.

[vMO]

is hyper-v "onhost" or "offhost" proxy used?