Monitoring system performance

[Arie Shoshani <arie@math.lbl.gov>]

As I mentioned in a previous message we plan to run dynamically
netlogger visualization of the logged to track the progress of 
queries.  Specifically, the tracking is planned to be done on the
status of each file.  The files will be colored according to the query
they belong to.

In addition, we are planning to monitor the performance of the system
as a whole, such as the cache utilization, tape drive starvation, etc.
Altough this was not originally planned for MDC2, we think this will
provide an understanding of the system resouce utilization.  We hope
to have this ready by MDC2 test period. 

Specifically, Luis and I developed the following plan.  We'll log a
few things that will help us understand the state of the query
service and caching status dynamically.  There are 3 measures of
interest:

1) Tape drive activity: is the tape drive busy or idle?
2) Cache activity: is the cache being used effectively by queries?
3) Query activity: how many queries are waiting for files and
   how many are processing?