[cgl_discussion] 01/09/2003 Poc Meeting Minutes

Peter Badovinatz tabmowzo at us.ibm.com
Thu Jan 9 15:58:37 PST 2003


"Randy.Dunlap" wrote:
> 
> On Thu, 9 Jan 2003, Lynch, Rusty wrote:
> 
> | ARS:
> |   - Ram to resend (or get Todd resend) previous Performance Co-Pilot
> |     project evaluation research
> Got it.

Re-reading it now.  Good job Todd!

> 
> |   - The rest of the team to take the time to read over the Performance
> |     Co-Pilot research

I vote in favor of Performance Co-Pilot (PCP).

> | _______________________________________________
> 
> Todd,
> 
> This seems to be a very good comparison of Resource Monitor and PCP.
> I do have a few comments and questions about it.
> 
> 1.  CGL requirement 4.1.4 Resource Monitor Data Persistence:
> OSDL CGL resource monitor framework shall provide the ability to capture and
> persist time stamped data. A set of APIs must be provided for an application
> to retrieve persisted historical data for analysis and display purposes.
> 
> Does RM or PCP address this?  If so, how?
> Maybe (probably) there's some lack of detail in this requirement,
> and that could be confusing me.

PCP has pmlogger which is supposed to allow archiving and replay of the
archive data, allowing you to thus source from the monitored systems as
data produced, or from the archive.  Not having actually done it myself
on a system, I cannot vouch that it truly works, or cleanly works, or
any such.

> 
> 2.  PCP seems to focus on system-level performance versus system-level
> resource monitoring.  Can you comment on this?
> 
> Can both PCP and RM be configured to send an email or other kind of alert
> for something trivial (but important) to someone when a disk partition
> reaches 95% full, e.g.?

PCP claims to be able to do this.  I'd agree that it's name and
documentation uses the phrase system performance, but it does this via
monitoring system resources.  It's how you lean your head to look at the
issue :o) 
> 
> 3.  PCP supports both local and remote monitoring of data.
> This is highly important IMO and a huge gap in RM.
> 
> 4.  The comparison does a good comparison of PCP and RM, but we need
> something that tells if/how they fulfull the CGL requirements also
> (4.1.2 thru 4.1.5) (as asked in #1 above).

I'm looking at the Req. doc now, and the only one that I wouldn't be
comfortable with saying is covered is the performance one (4.1.5)
because I've not the hands on experience with that.  If Todd is
comfortable there, then I would be too.  I think PCP well covers all of
the other points (with limitations and future work directions as
described by Todd.)

> 
> Thanks,
> --
> ~Randy

Peter
--
Peter R. Badovinatz aka 'Wombat' -- IBM Linux Technology Center
preferred: tabmowzo at us.ibm.com / alternate: wombat at us.ibm.com
These are my opinions and absolutely not official opinions of IBM, Corp.



More information about the cgl_discussion mailing list