[rabbitmq-discuss] In-memory Cache or Messages?

Marek Majkowski majek04 at gmail.com
Fri Dec 2 14:45:20 GMT 2011

On Thu, Dec 1, 2011 at 21:58, Bryan Burke <bryan.t.burke at gmail.com> wrote:
> I have an application that performs pretty hefty number crunching on
> large arrays.  Right now, there is an independent process that reads
> the data from disk and caches it in Redis, passing the key to the
> worker (basically pointer to data structure).  I'm not rewriting the
> task queue with RabbitMQ.  Would it be better to pass these arrays
> directly as messages or to stick with a scheme of passing the keys to
> the queue and keeping the arrays in the cache?

The clear answer is that there is no clear answer!

Do benchmarks. Frankly. If the arrays aren't too big you should be
better off passing them as part of messages. OTOH if your
client is badly written and slow, you may be better off asking Redis
about the data.

But do benchmarks, implement both and measure what behaves better
in your situation.


More information about the rabbitmq-discuss mailing list