Ivan Voras wrote:
On 18 February 2011 15:06, Hugo Silva <h...@barafranca.com> wrote:

Performance seems to be acceptable. There's a gotcha with PF, which someone
else mentioned in this list recently. One has to disable tcp.tso to get
decent throughput.

Disabling it enabled a colleague who is currently in Africa to go from
stalled..2KB/s sftp connections to 70KB/s.

I hope that 70 kB/s is due to him being in Africa and not Xen+FreeBSD
performance :) I'm thinking of using this on a moderately loaded web
server (cca 5 GB/day traffic).

Definitely - I can upload files (via sftp) to this VM at ~90% my home connection upstream, over a UDP VPN.

Also, a few weeks ago while setting this up, on local LAN tests it managed to push 400-600mbit/s (iperf) without a lot of tuning.

Back when I ran the iperf test PF wasn't enabled yet, so that might explain how it managed to get to that (altough I haven't tested the other poster's assertion that the performance problem only manifests with PF on)

In my experience NetBSD tends to perform better under Xen but that could be because HVM+PV vs PV. Also a NetBSD PV domU is limited to a single CPU, while with FreeBSD HVM+PV you can use many vcpus.

I haven't tested with high vcpu counts, but in this particular VM, it's been running happily with 2 VCPUs and no noticeable problems. No panics, no noticeable performance problems (after the tso fix) and certainly not the famous 'time went backwards' problem.
freebsd-xen@freebsd.org mailing list
To unsubscribe, send any mail to "freebsd-xen-unsubscr...@freebsd.org"

Reply via email to