You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@jmeter.apache.org by Srinims <sr...@aztec.soft.net> on 2006/09/07 15:41:42 UTC
Throughput
Hi ,
Just had a quick clarification.
Is the throughput value that is displayed in Jmeter the server side
throughput or the client side throughput ?
I noticed a very unusual case.
When I load test a server by running Jmeter on one box then there is a
value of throughput X.
Now when I simulate the same load and number of requests from 2 different
boxes the value of throughput is a lot different.
Now logically speaking for a given set of conditions the server throughput
should not change ( i.e the processing capability in terms
of requests/sec or transactions/sec should not vary much for that server).
Pls clarify if it is "SERVER SIDE " throughput that is being displayed .If
so how is it calculated ?
Rgds
srims
--
View this message in context: http://www.nabble.com/Throughput-tf2233234.html#a6190586
Sent from the JMeter - User forum at Nabble.com.
Re: Throughput
Posted by Marc Anthony Winoto <ma...@netregistry.com.au>.
From the manual:
"The throughput number represents the actual number of requests/minute the server handled. This
calculation includes any delays you added to your test and JMeter's own internal processing
time. The advantage of doing the calculation like this is that this number represents something
real - your server in fact handled that many requests per minute, and you can increase the
number of threads and/or decrease the delays to discover your server's maximum throughput.
Whereas if you made calculations that factored out delays and JMeter's processing, it would be
unclear what you could conclude from that number."
So it's the server side throughput. That is, the number of requests the server handled. Given
they way the measurements are taken, you'd potentially need to be very careful to get the same
results each time.
I myself get quite different times from the same machine sometimes.
Srinims wrote:
> Hi ,
>
> Just had a quick clarification.
> Is the throughput value that is displayed in Jmeter the server side
> throughput or the client side throughput ?
> I noticed a very unusual case.
> When I load test a server by running Jmeter on one box then there is a
> value of throughput X.
> Now when I simulate the same load and number of requests from 2 different
> boxes the value of throughput is a lot different.
> Now logically speaking for a given set of conditions the server throughput
> should not change ( i.e the processing capability in terms
> of requests/sec or transactions/sec should not vary much for that server).
> Pls clarify if it is "SERVER SIDE " throughput that is being displayed .If
> so how is it calculated ?
>
> Rgds
> srims
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
RE: Throughput
Posted by Alex Turner <Al...@Project-Network.com>.
I never really thought about it - I always assumed it was the rate at which data returned from the server - it certainly seems to be.
As to your point about 1 box vs 2, I doubt your logic will hold in reality. First off you will have to know what is the bottle neck. On my test system, for small static files TAG.net is only just waking up on a 2.6GHz P4 when JMeter is using all of a 3.6 P4! In this situation, you'd find that the load from two JMeter test boxes would be very much greater than from 1.
Then you have network bandwidth that will be different with three machines instead of 2. Ethernet will normally saturate at a lower throughput if there are more machines. This is because one card will never collide with its self.
Then you have the tcp/ip stack on the machines. This varies massively from one OS to another. However, if you are running without keepalive, it is very easy for the server machine or the client machine to run out of user sockets. This results in dropped connections if it is the server or the JMeter slowing down if it is the client (or a mix of both - it is a mess).
If you are running out of sockets on the client machines, then running two machines will help.
To get reasonable HTTP1.0 performance from a Windows box - for example - requires a bunch of registry hacks.
Hope that helps shed some light!
Cheers
AJ
Alexander J Turner Ph.D.
www.deployview.com
www.nerds-central.blogspot.com
www.project-network.com
-----Original Message-----
From: Srinims [mailto:srinivass@aztec.soft.net]
Sent: 07 September 2006 14:42
To: jmeter-user@jakarta.apache.org
Subject: Throughput
Hi ,
Just had a quick clarification.
Is the throughput value that is displayed in Jmeter the server side
throughput or the client side throughput ?
I noticed a very unusual case.
When I load test a server by running Jmeter on one box then there is a
value of throughput X.
Now when I simulate the same load and number of requests from 2 different
boxes the value of throughput is a lot different.
Now logically speaking for a given set of conditions the server throughput
should not change ( i.e the processing capability in terms
of requests/sec or transactions/sec should not vary much for that server).
Pls clarify if it is "SERVER SIDE " throughput that is being displayed .If
so how is it calculated ?
Rgds
srims
--
View this message in context: http://www.nabble.com/Throughput-tf2233234.html#a6190586
Sent from the JMeter - User forum at Nabble.com.
--
No virus found in this incoming message.
Checked by AVG Free Edition.
Version: 7.1.405 / Virus Database: 268.12.1/440 - Release Date: 06/09/2006
--
No virus found in this outgoing message.
Checked by AVG Free Edition.
Version: 7.1.405 / Virus Database: 268.12.1/440 - Release Date: 06/09/2006
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org