activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Vibe from Boston <djvibe_...@yahoo.com>
Subject My Benchmark stats
Date Sun, 18 Apr 2010 14:37:34 GMT

I will be soon posting the benchmark stats on my blog, but I wanted to post
it here.
all on one meaty machine, Intel I7 64bit Windows server, 8GB ram.

Client sends a message to nio based server [thru xsocket lib]. the nio
server also sends an acknoledge message [not JMS akg]
nio server then publishes to a Q. another jvm instance listens on this queue
and processes the message accordingly. 
at first, out of th box configuration for active mq, the whole thing ran for
1.4 mins for 50,000 messages.

then i tweaked, it ran in 20 seconds. its absolutely amazing
[25,000/second]. I really don't care how long the listener would take but i
do care how much the client took. [the whole intent of going to activemq is
to defer the processing for a later stage].

my next test is going to be putting both the nio server/active mq in EC2 and
call client from local. that would be the true test.

the tweaks are already well documented, but here it is anyway
<code>
alwaysSessionAsync="true" alwaysSyncSend="false" copyMessageOnSend="false"
disableTimeStampsByDefault="false" dispatchAsync="true" useAsyncSend="true"
</code>

i am assuming alwaysSyncSend and useAsyncSend are the same?


-- 
View this message in context: http://old.nabble.com/My-Benchmark-stats-tp28282606p28282606.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.


Mime
View raw message