How to randomize active robots

[perf tester]

Hi Alex,

Thanks for the reply. The second part of your reply has detailed
explanation for my question and possible solution. And it is very useful
feature.

For now, I will play with idle and busy period durations and see if
can achieve my requirements.

Thanks,
Krish.

On Wed, Mar 9, 2016 at 10:50 AM, Alex Rousskov <
rousskov at measurement-factory.com> wrote:

> On 03/08/2016 12:24 AM, perf tester wrote:
>
> > I would like randomize the active robots. For example, Out of 100
> > robots configured in the test, i can limit 5 robots to be active. But
> > the same robots are used for the entire test duration. I would like
> > the tool to pickup the next 5 robots from the remaining 95 robots
> > configured until all the robots are used 5 at a time.
>
> You would like Polygraph to "pickup the next 5 robots" _when_?
>
> If your requirements are flexible, and you do not need Polygraph to
> maintain exactly N active robots at any given time, then you may be able
> to achieve something similar using Robot sessions:
>
>
> http://www.web-polygraph.org/docs/reference/pgl/types.html#type:docs/reference/pgl/types/Session
>
> I have not tested it, but I suspect that by selecting the right
> combination of average idle and busy period durations while randomizing
> idle_period_duration, you may achieve a steady state where 5
> random/different robots out of 95 robots are active at most times.
>
>
> If you need more precise control and convenience, then Polygraph changes
> would be required. For example, one could add PGL knobs that would
> replace one active robot with another while maintaining a configured
> robot population level. Polygraph already has code to maintaining a
> configured robot population level, but it currently does not add or
> remove robots unless that level changes. This would be a generally
> useful feature IMO.
>
>
> HTH,
>
> Alex.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.web-polygraph.org/pipermail/users/attachments/20160309/9f5639f4/attachment.html>