Re: [rdiff-backup-users] Backup performance

[Dave Kempe]

Simon Hobson wrote:
> Is it normal for the target of a backup to show 100% wait-io during a 
> backup ?
>
>
> Both systems are guests on a Xen host with a 4core 2.4G processor and 
> plenty of ram (the target currently has over 1G free). rdiff-backup is 
> ver 1.2.1
I dunno about fixing rdiff-backup in this instance - strace will show 
you what it is doing pretty clearly.
However, as for your combination, I found that the network and I/O 
scheduler in Xen was a single pipeline and contention was terrible. We 
got terrible performance when we used network block devices with Xen, as 
the VMs would just sit in waitI/O all the time when accessing the 
network block devices (we tried AoE, NBD, iSCSI).
Also, use dstat in and out of the vm to analyse the throughput. I found 
that Xen just couldn't measure up when using network block devices as 
storage (primary storage or additional).
We ended up moving to OpenVZ and haven't looked back.

thanks

dave