You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@jmeter.apache.org by Manish Mathuria <mm...@infostretch.com> on 2005/08/24 20:01:28 UTC
RE: SPAM-LOW: Re: Test Infrastructure Requirements
For calculations sake, let us assume 0 think time and I can scale it up as I
increase the think times.
So at 0 think times I should be able to get 20-25 concurrent SOAP requests
for a 2.0 Ghz CPU? How much memory would I need with the 2.0 Ghz CPU? Will I
see CPU max out at that point or RAM? With think times of approx 30 seconds
is it safe to assume I will be able to get 50 users from 1 mc?
>you can actually calculate the theoritical max concurrent load the server
>can handle soap webservices without running any tests, since xml
>performance is linear.
The soap requests make the server hit several components (like DB, message
queues etc. ) which are non linear. And like all other systems soap
application server does not seem to be a bottleneck here.
----------------------------------------------------------------------------
---
Manish Mathuria
InfoStretch Corporation (www.infostretch.com)
(510) 673 6197 (Cell)
(408) 200 7450 (Work)
-----Original Message-----
From: Peter Lin [mailto:woolfel@gmail.com]
Sent: Wednesday, August 24, 2005 10:44 AM
To: JMeter Users List
Subject: SPAM-LOW: Re: Test Infrastructure Requirements
that depends on how much think time you have between requests. a 2ghz CPU
system should be able to simulate 20-25 concurrent soap requests due to XML
parser performance. that's the primary limitation.
you can actually calculate the theoritical max concurrent load the server
can handle soap webservices without running any tests, since xml performance
is linear.
peter
On 8/24/05, Manish Mathuria <mm...@infostretch.com> wrote:
>
> Hi
>
> I am looking to simulate about 300 concurrent users all making one or
> more SOAP calls. The SOAP calls will be grouped in more business centric
> scenarios and I will divide the users such that they are executing about 6
> different scenarios.
>
> My question is how should I go about planning the test hardware capacity
> (i.e. the machines I will need for JMeter peers generating the load - and
> not the server side infrastructure). I intend to use windows XP.
>
> Please advise
>
>
>
----------------------------------------------------------------------------
---
>
> Manish Mathuria
>
> InfoStretch Corporation (www.infostretch.com <http://www.infostretch.com>)
>
> (510) 673 6197 (Cell)
>
> (408) 200 7450 (Work)
>
> ** <http://www.infostretch.com/home/index.htm>
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
Re: SPAM-LOW: Re: Test Infrastructure Requirements
Posted by Peter Lin <wo...@gmail.com>.
I have an old article I wrote about tomcat performance. It's listed in
tomcat's resource page.
On 8/24/05, Manish Mathuria <mm...@infostretch.com> wrote:
> For calculations sake, let us assume 0 think time and I can scale it up as I
> increase the think times.
>
> So at 0 think times I should be able to get 20-25 concurrent SOAP requests
> for a 2.0 Ghz CPU? How much memory would I need with the 2.0 Ghz CPU? Will I
> see CPU max out at that point or RAM? With think times of approx 30 seconds
> is it safe to assume I will be able to get 50 users from 1 mc?
this is assuming there's atleast 512-1gb of RAM. xml parser
performance to the best of my knowledge is linear regardless of the
platform. by that I mean it doesn't matter if it's C++, Java or .NET.
I've run tests on 2.4 and 2.6ghz Pentium 4 CPU's. Generally, parser
performance for 2-2.6ghz will max out the CPU around 25 concurrent
parser processes. I wouldn't want to run for extended period of time
at 100% cpu utilization, so a more realistic number would be 75% CPU
usage. that basically means 18 concurrent. Once you know how long it
takes to respond, you can calculate the max req/sec your server can
handle.
lets say you've tuned the WS and they complete in 500ms. that means
the max through req/sec is around 32. JMeter can easily generate 32
req/sec with one 2.5ghz workstation. I hope that helps
peter
>
> >you can actually calculate the theoritical max concurrent load the server
> >can handle soap webservices without running any tests, since xml
> >performance is linear.
> The soap requests make the server hit several components (like DB, message
> queues etc. ) which are non linear. And like all other systems soap
> application server does not seem to be a bottleneck here.
> ----------------------------------------------------------------------------
> ---
> Manish Mathuria
> InfoStretch Corporation (www.infostretch.com)
> (510) 673 6197 (Cell)
> (408) 200 7450 (Work)
>
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org