[rabbitmq-discuss] Poor performance using a single RabbitMQ connection on high-latency networks

[Matthias Radestock]

Holger,

Holger Hoffstaette wrote:
> I'm not sure why the values have to be even smaller than e.g. Linux'
> vanilla defaults. Ideally - and given that Rabbit is mostly used on
> Linux/Unix - they should not be necessary at all, especially as they
> prevent any client-side settings from taking effect (I think - would
> have to check with wireshark).

Our sndbuf setting actually is the same as gets reported by Erlang by 
default on the systems I've tried - 16k. OTOH, the recbuf setting of 4k 
is significantly smaller than the 85k Erlang reports by default.

As I mentioned before, I tried to minimise per-connection memory usage. 
So the figures are the lowest I could find that showed good performance 
in my tests.

But yes, I'd rather stick to the defaults too. See below.

> The comment re. memory usage with a lot of connections is interesting
> though - is this really a problem even with epoll (aka kernel polling)?
> What would be a large number of connections - 100? 1000?

Depends what you are running on. A linksys slug - the smallest device 
we've run rabbit on - can quite happily handle thousands of connections, 
but not if they each take up >100k off memory.

OTOH, if you are running rabbit under extreme conditions then perhaps 
it's not too much to ask for some manual parameter tweaking.

So I think we should investigate sticking to the defaults, and let users 
tweak the settings when they need to. That's also what decided to do in 
the client libraries.


Matthias.