Overview questions

[James Hill <ev...@gmail.com>]

Hi list. I'll try and keep this short.

I've been using jmeter for a couple of weeks now and while I'm still a
newbie I find it very easy to use - once I got past a couple of
hiccups. I've been building a suite of load tests for a site in
development via proxy recording but the dev lead wants me to test the
pages "as a real world user would" whatever that means. I think what
he's getting at is for me to reassure him (and more importantly the
client) that jmeter will be loading up the server as a bunch of users
would i.e. making page requests that include all the html, css,
images, javascript. Unfortunately what I think his "real user" comment
is getting at is asking "Does jmeter access a site like a user would
i.e. clicking through the links, etc."

I'm sure it doesn't, but how do I go about telling him that jmeter
does in fact mimic a real world load i.e. pages being requested and
loaded like they would be by users clicking on links? I've been
reading doco but so far haven't come across something I can quote at
him. I think he's nervous jmeter won't actually do what I tell him it
does.

---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org

RE: Overview questions

[Matthew Coventon <ma...@iseinc.biz>]

I think testing real world usage is very important.  If your application
can't handle 100 users doing everyday tasks at a normal pace then testing
with 200 users doing resource intensive tasks at a very fast pace doesn't do
any good.

One thing you can point out to your dev lead is that you are load testing
the server, and the server doesn't see people clicking through links it sees
requests and generates responses.

As far as generating a realistic user simulation, I think JMeter has
everything you need.  Someone mentioned think times already (I think its
best to record them as Sebb mentioned ... the docs tell you how to record
them into a gaussian timer which is best).  Ramp-up times are important.
And another important thing is using thread groups and JMeter controllers to
simulate a real workload distribution.  For example, if you have an online
music store and your simulation had every user eventually making a purchase
so that the "browse music" section got just as much traffic as the "buy
music" section ... that would not be realistic.  You can use JMeter to have
some users that browse also buy and others just abandon the site.

I've learned a lot about performance testing from these articles:
http://www.perftestplus.com/pubs.htm.  There is a series of articles called
"User Experience, not Metrics" (about 2/3 down the page) that I have found
very useful.  I hope they help you too.

Regards,

matt.

-----Original Message-----
From: James Hill [mailto:everywherenet@gmail.com] 
Sent: Monday, March 19, 2007 6:24 PM
To: jmeter-user@jakarta.apache.org
Subject: Overview questions

Hi list. I'll try and keep this short.

I've been using jmeter for a couple of weeks now and while I'm still a
newbie I find it very easy to use - once I got past a couple of
hiccups. I've been building a suite of load tests for a site in
development via proxy recording but the dev lead wants me to test the
pages "as a real world user would" whatever that means. I think what
he's getting at is for me to reassure him (and more importantly the
client) that jmeter will be loading up the server as a bunch of users
would i.e. making page requests that include all the html, css,
images, javascript. Unfortunately what I think his "real user" comment
is getting at is asking "Does jmeter access a site like a user would
i.e. clicking through the links, etc."

I'm sure it doesn't, but how do I go about telling him that jmeter
does in fact mimic a real world load i.e. pages being requested and
loaded like they would be by users clicking on links? I've been
reading doco but so far haven't come across something I can quote at
him. I think he's nervous jmeter won't actually do what I tell him it
does.

---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org




---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org

Re: Overview questions

[sebb <se...@gmail.com>]

By the way, you can get JMeter to add a timer to each sampler - see
the documentation:

http://jakarta.apache.org/jmeter/usermanual/component_reference.html#HTTP_Proxy_Server

<quote>
It is also possible to have the proxy add timers to the recorded
script. To do this, create a timer directly within the HTTP Proxy
Server component. The proxy will place a copy of this timer into each
sample it records, or into the first sample of each group if you're
using grouping. This copy will then be scanned for occurences of
variable ${T} in its properties, and any such occurences will be
replaced by the time gap from the previous sampler recorded (in
milliseconds).
</quote>

On 19/03/07, James Hill <ev...@gmail.com> wrote:
> That might work. And explaining what jmeter does as a maximum load
> test rather than a user test (that's why we have functional tests,
> right?) could alter the perspective a little.
>
> On 3/20/07, Marc Anthony Winoto <ma...@netregistry.com.au> wrote:
> > How about inserting random timers between samplers? Would that be closer to a real human? I've
> > modified test with random timers of a few seconds to simulate "reading" and "mouse moving and
> > clicking" time. Just for experimentation.
> >
> > Also, I don't think the point is to mimic a real bunch of users. Rather, to simulate a load
> > _beyond_ what you expect. If your application can handle that, then real users are easy.
> >
> >
> >
> > James Hill wrote:
> > > Hi list. I'll try and keep this short.
> > >
> > > I've been using jmeter for a couple of weeks now and while I'm still a
> > > newbie I find it very easy to use - once I got past a couple of
> > > hiccups. I've been building a suite of load tests for a site in
> > > development via proxy recording but the dev lead wants me to test the
> > > pages "as a real world user would" whatever that means. I think what
> > > he's getting at is for me to reassure him (and more importantly the
> > > client) that jmeter will be loading up the server as a bunch of users
> > > would i.e. making page requests that include all the html, css,
> > > images, javascript. Unfortunately what I think his "real user" comment
> > > is getting at is asking "Does jmeter access a site like a user would
> > > i.e. clicking through the links, etc."
> > >
> > > I'm sure it doesn't, but how do I go about telling him that jmeter
> > > does in fact mimic a real world load i.e. pages being requested and
> > > loaded like they would be by users clicking on links? I've been
> > > reading doco but so far haven't come across something I can quote at
> > > him. I think he's nervous jmeter won't actually do what I tell him it
> > > does.
> > >
> > > ---------------------------------------------------------------------
> > > To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> > > For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
> > >
> > >
> >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> > For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
> >
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
>
>

---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org

Re: Overview questions

[James Hill <ev...@gmail.com>]

That might work. And explaining what jmeter does as a maximum load
test rather than a user test (that's why we have functional tests,
right?) could alter the perspective a little.

On 3/20/07, Marc Anthony Winoto <ma...@netregistry.com.au> wrote:
> How about inserting random timers between samplers? Would that be closer to a real human? I've
> modified test with random timers of a few seconds to simulate "reading" and "mouse moving and
> clicking" time. Just for experimentation.
>
> Also, I don't think the point is to mimic a real bunch of users. Rather, to simulate a load
> _beyond_ what you expect. If your application can handle that, then real users are easy.
>
>
>
> James Hill wrote:
> > Hi list. I'll try and keep this short.
> >
> > I've been using jmeter for a couple of weeks now and while I'm still a
> > newbie I find it very easy to use - once I got past a couple of
> > hiccups. I've been building a suite of load tests for a site in
> > development via proxy recording but the dev lead wants me to test the
> > pages "as a real world user would" whatever that means. I think what
> > he's getting at is for me to reassure him (and more importantly the
> > client) that jmeter will be loading up the server as a bunch of users
> > would i.e. making page requests that include all the html, css,
> > images, javascript. Unfortunately what I think his "real user" comment
> > is getting at is asking "Does jmeter access a site like a user would
> > i.e. clicking through the links, etc."
> >
> > I'm sure it doesn't, but how do I go about telling him that jmeter
> > does in fact mimic a real world load i.e. pages being requested and
> > loaded like they would be by users clicking on links? I've been
> > reading doco but so far haven't come across something I can quote at
> > him. I think he's nervous jmeter won't actually do what I tell him it
> > does.
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> > For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
> >
> >
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
>
>

---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org

Re: Overview questions

[Marc Anthony Winoto <ma...@netregistry.com.au>]

How about inserting random timers between samplers? Would that be closer to a real human? I've 
modified test with random timers of a few seconds to simulate "reading" and "mouse moving and 
clicking" time. Just for experimentation.

Also, I don't think the point is to mimic a real bunch of users. Rather, to simulate a load 
_beyond_ what you expect. If your application can handle that, then real users are easy.



James Hill wrote:
> Hi list. I'll try and keep this short.
> 
> I've been using jmeter for a couple of weeks now and while I'm still a
> newbie I find it very easy to use - once I got past a couple of
> hiccups. I've been building a suite of load tests for a site in
> development via proxy recording but the dev lead wants me to test the
> pages "as a real world user would" whatever that means. I think what
> he's getting at is for me to reassure him (and more importantly the
> client) that jmeter will be loading up the server as a bunch of users
> would i.e. making page requests that include all the html, css,
> images, javascript. Unfortunately what I think his "real user" comment
> is getting at is asking "Does jmeter access a site like a user would
> i.e. clicking through the links, etc."
> 
> I'm sure it doesn't, but how do I go about telling him that jmeter
> does in fact mimic a real world load i.e. pages being requested and
> loaded like they would be by users clicking on links? I've been
> reading doco but so far haven't come across something I can quote at
> him. I think he's nervous jmeter won't actually do what I tell him it
> does.
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
> 
> 


---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org