[rabbitmq-discuss] Odd performance testing result

[Michael Klishin]

On 24 Jan 2014, at 08:28, Huang Tonglang <tlhuang at 800best.com> wrote:

> During testing, I find the cpu us of machine-3 can reach to 90, but it is only about 50 in machine-2. It seems the message persistence stage in machine-2 is very slow.

If you have 50 or even 90 cores and want to achieve max throughput, use more queues (e.g. as many as you have cores).

> Can someone help me to figure out why.

vmstat and iostat can help with that.

MK

Software Engineer, Pivotal/RabbitMQ