Hello all,
I hate to contradict, but I noted figures of:
sync publish 15k msg/sec
tx size =2 48 msg/sec (thats right just 48 messages per sec)
tx size=1000 14k msg/sec
mentioned by Rajith in a previous post.
I seem to be getting something very different. I'm using Qpid 0.8 with C++
Broker and Java JMS Clients
With 1K message size I get
With no explicit sync (default behaviour) 21000 messages/sec
With sync_publish='all' 1420 messages/sec
With transacted session tx size = 1 1125 messages/sec
With transacted session tx size = 500 20750 messages/sec
With 50K message size I get
With no explicit sync (default behaviour) 1500 messages/sec
With sync_publish='all' 332 messages/sec
With transacted session tx size = 1 314 messages/sec
With transacted session tx size = 500 1550 messages/sec
I do wonder about the use of sync_publish though! When I use the connection
URL approach, e.g.
connectionfactory.ConnectionFactory =
amqp://guest:guest@clientid/test?brokerlist='tcp://localhost:5672'&sync_publish='all'
it behaves as expected, however if I do
-Dsync_publish=all
there is no effect......
I had a quick grep through the Java client code and I suspect a bug in
AMQConnection.java as this has a line that says:
"_syncPublish =
System.getProperty((ClientProperties.SYNC_ACK_PROP_NAME),_syncPublish);"
The SYNC_ACK_PROP_NAME looks wrong and I guess should really say
SYNC_PUBLISH_PROP_NAME.
I re-read one of Rajith's previous posts and note that "Therefore I
recommend you use -Dsync_ack=persistent" is mentioned, but given the mixture
of that and "-Dsync_publish={persistent|transient|all}. " mentioned in the
same post I wonder if Rajith has perhaps accidentally used the (non-working)
-Dsync_publish=all option with the test giving the result "sync publish
15k msg/sec"?? so that value may really be for the default mode?
I'm also really surprised at the 48 messages/second result with a tx size of
2. As I say above even with 50K messages I'm still getting 314 messages per
second with Java client and C++ broker and tx size of 1 (I'm running clients
and broker on the same box, so I'd expect it to be even faster if I ran my
clients on a separate box).
Regards,
fadams
--
View this message in context:
http://apache-qpid-users.2158936.n2.nabble.com/Qpid-CLIENT-ACKNOWLEDGE-tp6196226p6210606.html
Sent from the Apache Qpid users mailing list archive at Nabble.com.
---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project: http://qpid.apache.org
Use/Interact: mailto:[email protected]
