Mailing-List: contact users-help@activemq.apache.org; run by ezmlm
Precedence: bulk
Reply-To: users@activemq.apache.org
Received-SPF: pass (athena.apache.org: domain of rajdavies@gmail.com
 designates 64.233.182.184 as permitted sender)
DomainKey-Signature: a=rsa-sha1; c=nofws;
        d=gmail.com; s=gamma;
        h=message-id:from:to:in-reply-to:content-type:content-transfer-encoding:mime-version:subject:date:references:x-mailer;
        b=kCv8kPJnFgBb1GUjkJXeNU2HwmDk/Fv0Zh1SuL4ZyswQlSeayNNMgbI9IIhQFwPYKRC7rWRQR+EUmI/npwkyrnIH6SinyoBAC/M/KsYneLxQRv8MeaaT3TNG2GJqBuLSTTZzbFTZ1WtvudX27AGDwI6hF5iNg4bSzb6hcu0zrVc=
Message-Id: <8FD77796-3AF2-401B-B0EE-8FC9CEB11EAD@gmail.com>
From: Rob Davies <rajdavies@gmail.com>
To: users@activemq.apache.org
In-Reply-To: <17524607.post@talk.nabble.com>
Content-Type: text/plain; charset=US-ASCII; format=flowed; delsp=yes
Content-Transfer-Encoding: 7bit
Mime-Version: 1.0 (Apple Message framework v919.2)
Subject: Re: Why producer to a queue or durable topic consumes a lot of CPU?
Date: Thu, 29 May 2008 07:11:13 +0100
References: <17524607.post@talk.nabble.com>


On 29 May 2008, at 00:57, janylj wrote:

>
> I had a single broker, which uses the default configuration. Then I  
> had a
> Producer sending messages to a queue and a Consumer reading from the  
> queue.
> At this time, the CPU usage on the broker is about 60%-70%. Because  
> the
> broker is on a box with 2 Xeon processes each @ 1.86GHz, each  
> processor has
> more than 50% idle. Then I added another Producer and Consumer to  
> the queue,
> it drove the combined CPU usage over 100% and each processor only  
> had 30%
> idle. I think mostly Producer is taking CPU resource, adding more  
> Consumer
> doesn't seem to put much CPU load.
>
> Finally the message rate for one producer is about 4K message/ 
> second. While
> the rate when having three producers simultaneously only is 1.5K
> message/second each.
>
> The question is that whether it's expected from Activemq or  
> something I did
> wrong. Also does it imply that adding more producer does not increase
> throughput, if not decrease it?
>
> Thank you very much.
> -- 
> View this message in context: http://www.nabble.com/Why-producer-to-a-queue-or-durable-topic-consumes-a-lot-of-CPU--tp17524607s2354p17524607.html
> Sent from the ActiveMQ - User mailing list archive at Nabble.com.
>

There is a lot of work going on in the broker marshaling/unmarshaling  
messages, adding them to an index in store etc - you will hit a  
plateau where you won't increase overall throughput by adding more  
producers. From your example, the overall throughput has increased  
from 4k msg/sec to 4.5k msg/sec by using 3 producers instead of 1 - so  
you haven't quite reached that point yet - but your close.

We will be looking at increasing throughput and reducing cpu for  
version 6.0 - which will be due out later this year.


cheers,

Rob

http://open.iona.com/products/enterprise-activemq
http://rajdavies.blogspot.com/