Axis2 simple http server memory management

[Patrick Duflot <pd...@gmail.com>]

Hi list,

I did a long-run test with the axis2 simple http server and the echo sample
client invoking the echo service every 3 seconds.
I saw the memory usage going from 4MB to 30MB then back to 10MB and then
again to 40MB. Now that I stopped the client from polling the service,
memory usage is as low as 80KB.

NB : I am running Windows XP SP2

What is axis2 http server memory management policy ?
Can we have some control over it ?

A couple of related question :
How does it behaves with memory fragmentation ?
And the most importatnt one : Is Axis2 simple http server production ready ?

Thank you

Patrick

Re: Axis2 simple http server memory management

[Nandika Jayawardana <ja...@gmail.com>]

Usually, production deployment using axis2_http_server.exe is not
encouraged. We recommend using either Apache2 module or IIS module that
comes with axis2c for production deployment.

axis2_http_server.exe and axis2_tcp_server.exe manage memory by cleaning up
the allocated memory at the end of processing each request. There could be
some memory leaks which accumulate with the number of requests handled and
cause the growing memory usage issue.

You should be able to track the memory leaks using valgrind. On Windows
platform , you can use a tool like purify to check for memory leaks.

Regards
Nandika

On Wed, Jan 13, 2010 at 1:07 PM, Martin Benedix <mb...@plusserver.de>wrote:

> Patrick Duflot schrieb:
>
>  Hi list,
>>
>> I did a long-run test with the axis2 simple http server and the echo
>> sample client invoking the echo service every 3 seconds.
>> I saw the memory usage going from 4MB to 30MB then back to 10MB and then
>> again to 40MB. Now that I stopped the client from polling the service,
>> memory usage is as low as 80KB.
>>
>> NB : I am running Windows XP SP2
>>
>> What is axis2 http server memory management policy ?
>> Can we have some control over it ?
>>
>> A couple of related question :
>> How does it behaves with memory fragmentation ?
>> And the most importatnt one : Is Axis2 simple http server production ready
>> ?
>>
>> Thank you
>>
>> Patrick
>>
>>  Hi Patrick, Hi list.
>
> This are very important questions and I am also interested in the answers.
> So I wrote this reply to bring the topic up again.
> We are using the simple tcp server under linux and I experienced a growing
> memory consumption using the echo service and a saturation on a level
> at approximately 40 mb. I think that this is a result of the high watermark
> memory usage strategy of linux.
> I wanted to use valgrind on the simple tcp server and did not manage to get
> any output from the tool.
> What would be a good way to examine the memory usage behaviour of the
> tcp server? Has anyone succeeded in using valgrind on the tcp server?
> Any suggestion is highly welcome :)
>
> Thanks in advance,
>
> Martin Benedix
>



-- 
http://nandikajayawardana.blogspot.com/
WSO2 Inc: http://www.wso2.com

Re: Axis2 simple http server memory management

[Martin Benedix <mb...@plusserver.de>]

Patrick Duflot schrieb:
> Hi list,
>
> I did a long-run test with the axis2 simple http server and the echo 
> sample client invoking the echo service every 3 seconds.
> I saw the memory usage going from 4MB to 30MB then back to 10MB and 
> then again to 40MB. Now that I stopped the client from polling the 
> service, memory usage is as low as 80KB.
>
> NB : I am running Windows XP SP2
>
> What is axis2 http server memory management policy ?
> Can we have some control over it ?
>
> A couple of related question :
> How does it behaves with memory fragmentation ?
> And the most importatnt one : Is Axis2 simple http server production 
> ready ?
>
> Thank you
>
> Patrick
>
Hi Patrick, Hi list.

This are very important questions and I am also interested in the answers.
So I wrote this reply to bring the topic up again.
We are using the simple tcp server under linux and I experienced a growing
memory consumption using the echo service and a saturation on a level
at approximately 40 mb. I think that this is a result of the high watermark
memory usage strategy of linux.
I wanted to use valgrind on the simple tcp server and did not manage to get
any output from the tool.
What would be a good way to examine the memory usage behaviour of the
tcp server? Has anyone succeeded in using valgrind on the tcp server?
Any suggestion is highly welcome :)

Thanks in advance,

Martin Benedix