Re: [jmob] Probes for CLIF

[Bruno Dillenseger <bruno.dillenseger@xxxxxxxxxxxxxxxxxxxx>]

Very interesting.

Petr, would it be a problem if this code is adapted and put somewhere in 
ObjectWeb GForge, typically inside the LeWYS project? I'm just wondering 
about license issues, since it would be more friendly for users to ship 
CLIF with LeWYS and a bunch of probes.
Emmanuel, what is or will be the license for LeWYS?

I agree with Petr that some probes will require a native implementation. 
I see as a key requirement to be able to deal with OS-dependent probes, 
including probes based on native libraries. I also agree with Emmanuel 
that separated probes are better, even though I also think that it 
should be possible to have a single probe watching more than one 
resource (in some cases, for practical or efficiency reasons). In my 
understanding, this can be considered as already supported by LeWYS 
since a probe may provide an arbitrary variety of values (if I remember 
well). Am I wrong?

In order to have a better view of how we may use all this with CLIF, I 
would first need more information about the fractalized LeWYS... (but 
I'm not in a hurry - I have plenty of other things to do about CLIF!)

Thanks to both of you for giving CLIF a chance to quickly get a very 
useful probe infrastructure.

Best regards,
-- Bruno.

Petr Tuma a écrit:

> Hello Emmanuel,
>
> > Thanks a lot for the links. We are currently implementing 100% Java 
> > probes in LeWYS but we will certainly compare the performance with a 
> > JNI implementations like the one you have in Xampler. In this second 
> > step, we will certainly reuse the Xampler code but split the 
> > different probes in separate classes for an enhanced modularity.
>
>
> Just on the side, the Xampler probe code has been tested also on 
> Solaris and Windows NT/2000/XP, besides Linux. Some of the probe data 
> on these platforms cannot be obtained without JNI (specialized 
> syscalls are needed for that), so the 100% Java version will probably 
> not be possible.
>
> The old problem of some of the probes needing too much CPU time to 
> gather data remains, but especially on closed source platforms it is 
> hard to remedy that :-( ...
>
> Regards, Petr