Hi Sitaram,
----- Original Message -----
> Hello Community,
> I am Sitaram, and I am interested to contribute to Performance Co-Pilot. I
> read the the ideas listed on the GSoC Ideas page and found these two ideas
> interesting.
> 1. Performance metric extraction using Go language agents
> 2. PCP versions of console tools
> I read the reply to all of the previous discussions on this mailing list and
> read the programmers guide for pcp mentioned in one of the reply.
> I understood the basic concepts used in pcp like pcp daemon, domain agents,
> collectors and metrics and I have built pcp from source using the
> instructions provided in the source ( Thanks to Nathan for helping with the
> errors) and tried few commands to test their working as explained on the pcp
> website. I also ran some QA tests (not the complete test suite though).
> Lukas has explained about first task in one of the replies earlier and I
> would like to know more about the second task.
>
Sounds like you are picking it up quickly, that's great. The second task there
is about writing new performance analysis tools to display performance data in
a variety of ways, using the PCP client APIs.
In the case of the htop tool, it would involve working on making htop able to
consume data from PCP using the PMAPI (instead of, or in addition to, directly
accessing the machine it is running on). http://hisham.hm/htop/ - all C code.
Once implemented, you would then work with the htop authors/maintainers to see
if the new code could be merged into the master htop git tree (as a bonus).
For the other, less complex tools - like pidstat - the aim is to write python
scripts to implement the same functionality as the original tool. Again, that
involves accessing the metric values using the PMAPI (as you read about in the
"PCP Programmers Guide") such that the data is accessible from pmcd(1) - either
remotely or locally - and also via recorded PCP archives.
cheers.
--
Nathan
|