Extrapolating performance data to a customer base

[Ivan Rancati <iv...@sharpmind.de>]

I apologize for the slightly off topic post, but I guess it's a topic that
performance testers often encounter.

In general, the issue is: after running some tests, I now know that a Web
server farm can serve x concurrent customers in an average of y seconds per
customer.

What is a reasonable formula to extrapolate this number to a customer base
that can be served by this web farm without degradation of performance?

I guess there are some general assumptions, such as:
- there are less users at night
- there are less users during weekends
- the load can be assumed to be uniformely spread during the peak times
- ...

Any actual statistics on how the customer requests are spread and/or
practical examples of how other people calculated this would be appreciated.

thanks
Ivan


-- 
View this message in context: http://www.nabble.com/Extrapolating-performance-data-to-a-customer-base-tp20075112p20075112.html
Sent from the JMeter - User mailing list archive at Nabble.com.


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

Re: Extrapolating performance data to a customer base

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

On 20/10/2008, Ivan Rancati <iv...@sharpmind.de> wrote:
>
>  I apologize for the slightly off topic post, but I guess it's a topic that
>  performance testers often encounter.
>
>  In general, the issue is: after running some tests, I now know that a Web
>  server farm can serve x concurrent customers in an average of y seconds per
>  customer.

The maximum response times are also very important for a human user.
What happens when the load increases? Does the service degrade gracefully?

>  What is a reasonable formula to extrapolate this number to a customer base
>  that can be served by this web farm without degradation of performance?

This will depend an awful lot on what the application is used for.

For example - do customers need to use it every day?
Are there special events that might cause a sudden rise in traffic?
Do you have control over such events (e.g. a special promotion)?

>  I guess there are some general assumptions, such as:
>  - there are less users at night

Night depends on where the customer is located.

>  - there are less users during weekends

Likewise - for some cultures Sunday is a day of work.

>  - the load can be assumed to be uniformely spread during the peak times

That is very unlikely to be the case, but you might be able to put
some min and max bounds on the traffic.

>  - ...
>
>  Any actual statistics on how the customer requests are spread and/or
>  practical examples of how other people calculated this would be appreciated.
>

If you have access to server logs for the actual (or a similar)
application, then that would provide a lot of useful data.

>  thanks
>  Ivan
>
>
>
>  --
>  View this message in context: http://www.nabble.com/Extrapolating-performance-data-to-a-customer-base-tp20075112p20075112.html
>  Sent from the JMeter - User mailing list archive at Nabble.com.
>
>
>  ---------------------------------------------------------------------
>  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