default elevator=noop for virtio block devices?
Hi folks,
would it make sense to make elevator=noop the default for virtio block devices? Or would you recommend to set this on the kvm server instead?
Any helpful comment would be highly appreciated
Harri
Quoted 6 lines · Harald Dunkel
Hi folks,
would it make sense to make elevator=noop the default for virtio block devices? Or would you recommend to set this on the kvm server instead?
I think leaving the defaults is best. The elevator on the guest serves to schedule I/O among processes in the guest, and the elevator on the host partitions I/O among the guests.
Signature · 1 line
error compiling committee.c: too many arguments to function
Quoted 4 lines · Avi Kivity
Quoted 3 lines · Harald Dunkel
Hi folks,
Quoted 8 lines
Quoted 4 lines
would it make sense to make elevator=noop the default for virtio block devices? Or would you recommend to set this on the kvm server instead?
I think leaving the defaults is best. The elevator on the guest serves to schedule I/O among processes in the guest, and the elevator on the host partitions I/O among the guests.
It depends on the workload. Khoa has seen cases where CFQ does not scale with multi-threaded workloads and deadline is preferred. But it's not one-size-fits-all, it depends on your workload and requires benchmarking.
Stefan
Stefan Hajnoczi <[email protected]> wrote on 03/09/2011 05:12:07 AM:
Quoted 12 lines · Avi Kivity
Quoted 10 lines · Harald Dunkel
Quoted 7 lines
Hi folks,
would it make sense to make elevator=noop the default for virtio block devices? Or would you recommend to set this on the kvm server instead?
I think leaving the defaults is best. The elevator on the guest serves
to
Quoted 7 lines
Quoted 2 lines
schedule I/O among processes in the guest, and the elevator on the host partitions I/O among the guests.
It depends on the workload. Khoa has seen cases where CFQ does not scale with multi-threaded workloads and deadline is preferred. But it's not one-size-fits-all, it depends on your workload and requires benchmarking.
Yes, for I/O-intensive workloads on fast storage systems, I've found that the deadline scheduler works best.
Thanks,
-Khoa