Matthias,<br><br>as you will see in the previous message, I tested the two settings for exchanges (with x-lvc / with direct) : the difference is only 10% oc CPU load and is still at 120% with direct exchanges. According to you, would it be possible to control the producer flow ? In fact a ckeck could be done on the publisher-side prior to publish any message : if the routeId belongs to a list of required routeIds publish it otherwise no.<br>
<br>the problem is to have a routeIds list uptodate which requires :<br>- to know the all list of bindings at pblisher startup<br>- to be notify of each new binding when running.<br><br>Would this be possible ?<br>Thanks for your help.<br>
Romain<br><br><div class="gmail_quote">2010/10/8 Matthias Radestock <span dir="ltr"><<a href="mailto:matthias@rabbitmq.com">matthias@rabbitmq.com</a>></span><br><blockquote class="gmail_quote" style="margin: 0pt 0pt 0pt 0.8ex; border-left: 1px solid rgb(204, 204, 204); padding-left: 1ex;">
Romain,<div class="im"><br>
<br>
RomainC wrote:<br>
<blockquote class="gmail_quote" style="margin: 0pt 0pt 0pt 0.8ex; border-left: 1px solid rgb(204, 204, 204); padding-left: 1ex;">
10 exchanges (with x-lvc plugin that stores in mnesia database the<br>
last value for each message)<br>
</blockquote>
<br></div>
Ah, there is your explanation for the high CPU load then. The lvc plugin has not been designed/tuned for high performance. It does a transactional mnesia write per publish, which is hugely expensive.<br><font color="#888888">
<br>
<br>
Matthias.<br>
</font></blockquote></div><br>