[zeromq-dev] Best practice

Martin Sustrik sustrik at fastmq.com
Thu Apr 9 23:39:02 CEST 2009


Todd,

 > Impressive, your queue beat my queue by a significant margin.  Here 
are the numbers on a pretty weak box

Nice, here's the same thing converted into nanoseconds required to 
process one message:

 > Raw-No queue in between- 625586.487332 Msg/sec

1599 ns/msg

 > 0MQ-1 source- 590318.772137 Msg/sec

1694 ns/msg - overhead of 95 nanoseconds

 >  My Old queue-433369.447454 Msg/sec

2308 ns/msg - overhead of 709 nanoseconds

The boost tends to be more dramatic on multicore boxes than on single 
core ones. So, it you've tested it on a single-core box, you'll probably 
get even better result on a multicore one.

Martin



More information about the zeromq-dev mailing list