Host-based backup of VMware vSphere VMs.
Post Reply
WPHIT
Novice
Posts: 3
Liked: never
Joined: May 02, 2012 8:47 am
Full Name: Wirral Partnership Homes LTD
Contact:

Slow throughput when running a reverse incremental

Post by WPHIT »

HI,

Hopefully someone can shed some light on this issue (I have logged it with Veeam) but I thought i'd ask here too.

I'm running a full backup of all my servers. This backup runs over the LAN (no direct SAN etc..).

When I do my full backup, it runs pretty fast (about 50Mb/s) read on each server disk.

When I do my reverse incrememntal, this drops right down to 5-7Mb/s meaning that if I have a server that gets around 100GB of changes in a day, it takes around 5 hours run the incremental.

If it's any help the bottleneck detection points to "Target" when doing a reverse incremental and "Source" when doing a full backup.

I've tried using another Backup Proxy but that didnt improve things.

Please help!

Andy
foggy
Veeam Software
Posts: 21138
Liked: 2141 times
Joined: Jul 11, 2011 10:22 am
Full Name: Alexander Fogelson
Contact:

Re: Slow throughput when running a reverse incremental

Post by foggy »

Hello, Andy!

What kind of target storage do you use? Seems like it is not designed for random I/O's which are inherent in reverse incremental mode (unlike sequential writes during the full job run).

Thanks.
WPHIT
Novice
Posts: 3
Liked: never
Joined: May 02, 2012 8:47 am
Full Name: Wirral Partnership Homes LTD
Contact:

Re: Slow throughput when running a reverse incremental

Post by WPHIT »

Hi,

We are using a PowerVault MD1200. This is connected via SAS directly to our Physical Veeam Server. Its formatted RAID 5 with 1 hot spare.

When running a full backup, the throughput is awsome. Could you please explain why Random I/O is different?

Thanks,
Andy
foggy
Veeam Software
Posts: 21138
Liked: 2141 times
Joined: Jul 11, 2011 10:22 am
Full Name: Alexander Fogelson
Contact:

Re: Slow throughput when running a reverse incremental

Post by foggy »

It puts much more heavier load on the target storage. Sequential I/O is typical to reads and writes of large files and requires processing of one block after another located next to it. Disk drive head has to move a little to read/write those blocks. As opposed to random I/O, which requires large numbers of seeks and rotations, and hence is much slower. Add to this the necessity to write new blocks and move old blocks to the rollback (3x number of I/O operations comparing to active full) and you get much slower speed of backup. With 100GB change rate per VM you are likely to have incremental job runs longer than the full ones. Btw, here is another good reading on that.
Post Reply

Who is online

Users browsing this forum: Gregor and 62 guests