activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Gary Tully <gary.tu...@gmail.com>
Subject Re: How does prefetch work with client acknoledgements?
Date Mon, 09 May 2011 09:30:04 GMT
yes, the idea of prefetch is to deliver another message on or before
an ack such that it is available on the client before it is needed.
Some deviation is expected because a fast consumer will be able to ack
messages faster than the broker can dispatch them because it is doing
dispatch in batches of 1 when the prefetch is 1.

This is why we use a larger prefetch by default.

With prefetch=0, a client pulls a single message from the broker,
there is no prefetch but there is a round trip to the broker for each
pull request, so while the mean will be higher, the deviation will be
less.

To get the lowest mean and deviation and understand what is happening
further, preload with X messages and set the prefetch to X/N where N
is the number of consumers.

see: http://activemq.apache.org/what-is-the-prefetch-limit-for.html

On 4 May 2011 16:31, Josh Carlson <jcarlson@e-dialog.com> wrote:
> If I have prefetch set to 1 and I retrieve a message, is another one dispatched before
it is ack'd? I'm trying to scale the number of consumers. My benchmark test pre-produces 50
messages for each consumer (all in one queue). Then sets N consumers active and I measure
the time it takes to pull the message of the queue.
>
> In my benchmarking I've found a very nice MEAN time to retrieve a message at 2ms, regardless
of the number of consumers. However, the deviation is very bad. My hypothesis is that some
consumers are getting stuck not able to pull messages off the queue because each consumer
can actually be dispatched 2 messages. Also the faster the consumer the worse the deviation.
>



-- 
http://blog.garytully.com
http://fusesource.com

Mime
View raw message