Michael S. Tsirkin
2009-Oct-26 18:48 UTC
performance regression in virtio-net in 2.6.32-rc4
Hi! I noticed a performance regression in virtio net: going from 2.6.31 to 2.6.32-rc4 I see this, for guest to host communication: [mst at tuck ~]$ ssh robin sh streamtest1 TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 11.0.0.3 (11.0.0.3) port 0 AF_INET : demo Recv Send Send Socket Socket Message Elapsed Size Size Size Time Throughput bytes bytes bytes secs. 10^6bits/sec 87380 16384 16384 10.20 7806.48 [mst at tuck ~]$ ssh robin sh streamtest1 TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 11.0.0.3 (11.0.0.3) port 0 AF_INET : demo Recv Send Send Socket Socket Message Elapsed Size Size Size Time Throughput bytes bytes bytes secs. 10^6bits/sec 87380 16384 16384 10.00 6814.60 Note: I had to revert 48925e372f04f5e35fec6269127c62b2c71ab794, and I applied a patch virtio-pci: fix per-vq MSI-X request logic which fixes a bug introduced by f68d24082e22ccee3077d11aeb6dc5354f0ca7f1. Any tips on debugging this? -- MST
On 10/26/2009 08:48 PM, Michael S. Tsirkin wrote:> Hi! > I noticed a performance regression in virtio net: going from > 2.6.31 to 2.6.32-rc4 I see this, for guest to host communication: > > Any tips on debugging this? >Lacking better advice, a bisect can help as a last resort. 'git bisect start -- drivers/net drivers/virtio' will probably find it fastest. -- error compiling committee.c: too many arguments to function
On Tue, 27 Oct 2009 05:18:35 am Michael S. Tsirkin wrote:> Hi! > I noticed a performance regression in virtio net: going from > 2.6.31 to 2.6.32-rc4 I see this, for guest to host communication:...> Size Size Size Time Throughput > bytes bytes bytes secs. 10^6bits/sec > > 87380 16384 16384 10.20 7806.48...> 87380 16384 16384 10.00 6814.60Hmm, needs a bisect I'd say. Thanks, Rusty.