[Rhm-users] RHM performance
Andy Grove
andy at codesuccess.com
Tue Dec 11 13:46:37 UTC 2007
I was comparing qpid java broker from the M2 release with RHM C++ broker from the MRG beta.
Thanks.
Sent from my BlackBerry® wireless device
-----Original Message-----
From: Alan Conway <aconway at redhat.com>
Date: Tue, 11 Dec 2007 08:39:02
To:Gordon Sim <gsim at redhat.com>
Cc:Andy Grove <andy at codesuccess.com>, rhm-users at redhat.com
Subject: Re: [Rhm-users] RHM performance
Gordon Sim wrote:
> Andy Grove wrote:
>> The client that starts with a single thread and then adds a new thread
>> every 5 seconds. Each thread continuously publishes a fixed size
>> message to one of 50 topics at random. Throughput is measured on the
>> client for each 5 second duration (the cost of setting up each thread
>> and connecting to JMS is excluded from the throughput measurement).
>
> I suspect a large part of the difference between the measurements for
> the two systems may be attributable to the fact that publish is
> synchronous for durable messages in 0-10 (the next message is not sent
> until the broker acknowledges the previous one) whereas in 0-8 there is
> no confirmation of the publication (the fact that send() returns does
> not necessarily imply that the broker has even received it, let alone
> stored it).
>
From my reading of Andy's mail, I thought he was comparing the Qpid
trunk Java broker with RHM (essentially the Qpid trunk C++ broker) using
the same Java client for both. In that case both tests are using 0-10.
Andy can you confirm or deny my understanding?
I also suspect that different ack behavior or something of that sort is
likely to be the problem - something that got implemented/interpreted
differently in the two brokers and shows up in this test case
particularly for some reason.
Cheers,
Alan.
More information about the Rhm-users
mailing list