On Sunday 11 April 2004 12:43 pm, Tom Shilson wrote:

> I would expect to use a tool like RMFPMS only to take a baseline and then
> in problem situations. I may install it everywhere but not run it at all on
> most Linuxes.  If RMFPMS says that process A is using 50% of the CPU and
> process B is using 10%, I would assume that A is a lot busier that B, even
> though the exact numbers may be wrong.  Is this a valid assumption?

Well, it would give you a general idea.  The advantage of running an agent
inside the Linux guest is that it gives you visibility into the way the Linux
scheduler is allocating resources, in it's own terms of processes and
threads. RMFPM doesn't do a very good job of discriminating the internal
Linux structures -- no better or worse than 'top' - because it's operating
from the same bogus input. It (the agent) may give you a general sense of
what's happening, but probably not with enough precision to be really helpful
over all.

> What do you think of using SNMP to collect day-to-day data?

It works, as far as it goes.  The SMI MIB-II defaults are more oriented to
network devices than systems, but if you have MIBs and instrumentation to
populate them, it's as valid a method as any other.  SNMP is also a lot
lighter-weight than RMF-PM (no XML transcoding required), so the numbers are
less skewed.

On the other hand, the data source is still suspect in that it's pulling it's
data from Linux's point of view. You'll need to correlate both Linux and VM
results for the same time period to get a true picture.

-- db

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to