• SPEForums need your contributions to build up a strong repository of performance engineering resources.

A graph of throughput vs increasing vusers. Would anyone have an...

P

Peter Bird

Guest
A graph of throughput vs increasing vusers. Would anyone have an explanation for the throughout behavior?
 
I suspect caching on a number of components and pre-fetching of data. I just wondered if anyone else had observed something similar.
 
Do you have a fixed iteration interval which matches your ramp step up?
 
May be ur user ramup was shown effect on throughput graph , coz if u can see a sudden spike occurred every time during ramup but once test went into steady state throughput become normal. During stress model test hicups normal in Response time graph n throughput graph but in ur graph it was occurred due to ramup only. Try to give gap between user to user during ramup and test it. I believe if u give time gap for every user during ramup throughput and response time were normal.
 
The iteration interval is not fixed. If i change the ramp up interval the result is the same.
 
My first guess is you have a single threaded process early on in your process. IE login or SSO is a common problem area for that!. This is indicated by the fact that the delay bump is consistent with the number of newly launched vusers and not total in flight then recovery. You should insert several more transactions in your script as well to cover where the application infrastructure pieces get utilized. I would be you will quickly find the offending transaction.
 
It's a tru-client script running in Firefox. SSO has been disabled for these tests.
 
Petar's right, this could be a problem with pacing. Or it could be that as users login for the first time, they download all the images, CSS, JS and other cacheable content. If they're still logged in, actively "transacting" with the AUT and the throughout is falling, perhaps the TruClient scripts are behaving like real users and not downloading content that they have already cached.
 
There should be no caching apart from browser caching. However the script is set to "Simulate new user on each iteration".
 
It looks when you ramp up, the init section of your script has heavy data bytes transfer, when as once the ramp is complete, your actions step begins which doesn't pull so much data. Check your scripts run time setting. Are you planning to login only once with the user and reiterate action section, or you plan to create new session for every single iteration.
 
Basically, “Throughput” is the amount of transactions produced over time during a test. It’s also expressed as the amount of capacity that a website or application can handle. Also before starting a performance test it is common to have a throughput goal that the application needs to be able to handle a specific number of request per hr.
 
All your scenarios are going in pacing mode at the same time.
 
Either tweak the ramp up pattern or the pacing in scripts
 
You must log in or register to reply here.
Back
Top