How is the average worked out exactly?

Hi,

The Average in the Statistics Component is a running average of the data from the last x minutes, where x is configurable in the Settings dialog under "History (min)".

In the coming release of loadUI 1.5 we have a lot more options for getting statistics from tests, so keep your eyes out for that!

Regards,
Dain
eviware.com

So if x is set to 10, do the average entries in the table log for minutes 1-9 represent whatever data was available up to that point? So after 5 mins the average is of the data available up to 5 mins and 8 mins data is of data available up to 8 mins. At 11 mins, minute 1 gets dropped from the stats?

Hi, I also have a question about average time.

when I got a result file as PDF, "Web page runner" and "Statistics" both have "avg". However, both are different.
I wonder which one is more reliable.

My project is very simple. Only web runner, trigger, and statistics. I ran it via command line and get a pdf file.
Could you explain about this?

ojo0608 wrote:
Hi, I also have a question about average time. when I got a result file as PDF, "Web page runner" and "Statistics" both have "avg". However, both are different. I wonder which one is more reliable. My project is very simple. Only web runner, trigger, and statistics. I ran it via command line and get a pdf file. Could you explain about this?

Hi,

The Statistics Component's metrics are calculated from values gathered during the last minute (per default, this can be changed in its settings).

However, we recommend everyone to use the Statistics Workbench instead as it fully replaces the Statistics Component. Read more about this here.

Regards

Henrik
SmartBear Software

Ok great so now I understand that each line in the stats table represents the previous minutes data. If I wanted to make this a little more accurate can i put 0.5 mins? or do I put 30 secs or is 1min the lowest value I can have?

Also what do the other column represent such as
TPS - I know this is threads per sec but what does it actually represent and whats the diff between that and Ave-Tps

AndyHughes wrote:
Ok great so now I understand that each line in the stats table represents the previous minutes data. If I wanted to make this a little more accurate can i put 0.5 mins? or do I put 30 secs or is 1min the lowest value I can have?

It's the lowest value that you can have. The Statistics Workbench though calculate averages for each 1 second, 6 seconds, 10 minutes, 2 hours and 12 hours for the whole execution and stores that into a database, that allows you to zoom, scroll and compare results.

The Statistics Workbench can also show long names for its statistics and have their exact definition better documented (at loadUI.org/Getting-results ).

AndyHughes wrote:
Also what do the other column represent such as TPS - I know this is threads per sec but what does it actually represent and whats the diff between that and Ave-Tps

TPS actually stands for Transactions Per Second (as documented for the Statistics Workbench here). That is: How many transactions was completed (basically how many responses that was received) per second. Average TPS is just the average of this over a period of time (in the case of the Statistics component, down to 1 minute).

One of the loadUI team members wrote a blog post that discusses why and when TPS is very important.

Regards

Henrik
SmartBear Software


PS. Sorry for pushing the Statistics Workbench all the time, but we really think that it's better than the Statistics Component and are probably going to remove that component in future releases to avoid confusion (but making it available as a custom component).

So to get this absolutely straight then, if this is my result data....


bearing in mind we said that each row represents the previous minutes data, we cant also say that Tps is the number of transactions in the last second, or is this the exception to the rule?

So my data which shows avg-tps of 82 in the first row represents everything that is available up to that point (which is only 200ms since I get a row per 200ms as set in the settings). So why isn't it the same as Tps? Tps says 72.7. Or am I asking too much of it?

It doesnt seem to me to be useful to have columns like Avg-Tps and Avg (which im presuming is the avg response time) covering the last minute of a test when there is a row per 200ms. The averages aren't useful if the load is being ramped up over this minute.

Maybe time to move to the Stats Workbench

That TPS and Avg-TPS diff sure looks strange.

However, since we have already deprecated that component internally, we will not have a look at it. However, if you find any non-expected results in the Statistics Workbench, we would love to hear and fix them Hope you understand.

/Henrik

yeah i think stats workbench is defo the way forward. Getting stats at 1 second intervals that relate to that 1 second is much better than getting stats over a whole minute, particularly if the load is varying a lot during the test and particularly if the test itself is only run over a few mins.
Cheers