[rabbitmq-discuss] High(er) latency with 1.5.1?

[Holger Hoffstätte]

Hi,

I was re-running some of the latency tests with 1.5.1 again and noticed
that the latency is much higher than before, even with TCP_NODELAY.

Everything on Windows, running MulticastMain with -r 10 -s 100 -i 3 -h
localhost gives:

recving rate: 10 msg/s, min/avg/max latency: 1062/1237/1798 microseconds
recving rate: 10 msg/s, min/avg/max latency: 1047/1228/1764 microseconds
recving rate: 10 msg/s, min/avg/max latency: 1000/1245/1666 microseconds
recving rate: 10 msg/s, min/avg/max latency: 1042/1227/1756 microseconds
recving rate: 10 msg/s, min/avg/max latency: 1014/1202/1807 microseconds
recving rate: 10 msg/s, min/avg/max latency: 1083/1250/1555 microseconds
recving rate: 10 msg/s, min/avg/max latency: 1091/1242/1862 microseconds
recving rate: 10 msg/s, min/avg/max latency: 1035/1166/1373 microseconds
recving rate: 10 msg/s, min/avg/max latency: 993/1209/1511 microseconds
recving rate: 10 msg/s, min/avg/max latency: 955/1151/1476 microseconds
recving rate: 10 msg/s, min/avg/max latency: 974/1173/1777 microseconds
recving rate: 10 msg/s, min/avg/max latency: 974/1122/1655 microseconds
recving rate: 10 msg/s, min/avg/max latency: 969/1351/7688 microseconds
recving rate: 10 msg/s, min/avg/max latency: 972/1175/1463 microseconds
recving rate: 10 msg/s, min/avg/max latency: 1021/1179/1679 microseconds
recving rate: 10 msg/s, min/avg/max latency: 993/1176/1811 microseconds

Linux is the same as far as I can tell - almost a factor of 3 higher than
before (~250-300us after we fixed Nagle). Could this be a consequence of
the internal scalability changes done for 1.5.0? Or maybe just a
regression in the test - I think I remember seeing some changes to the
timing measurements in the changelogs.
Any ideas?

cheers
Holger