[Gllug] Historical performance reporting

John Hearns hearnsj at googlemail.com
Fri Mar 27 11:55:35 UTC 2009


2009/3/27 Stephen Nelson-Smith <sanelson at gmail.com>:
> What's currently recommended?  I have a dozen or so machines, and I
> want to look at trends for a number of metrics, and present them via a
> web interface to intelligent, non-technical stakeholders.

In the HPC world, Ganglia is the tool of choice for this.  http://ganglia.info/
It multicasts a range of metrics (you can add your own) to the central
Ganglia server, which stores the metrics in an RRDtool database.
There's really no setup needed - just install it and off you go. You
don't need a big cluster of machines - it would happily work with one
machine, and it produces the trend graphs you want out of the box.

I'm looking at the Ganglia webpage for one of my machines right now.
You can graph periods of one hour / day / week / month /year
Default metrics for the graphs are load, running processes, memory
(used, shared, cached, swapped,buffered)
cpu percentages (user, nice, system, wait, idle)  network (packetrs or
bytes in and out)
There are 'constant metrics' too - boottime, OS release, number of CPUs etc.






Alternatives you could look at are sysstat/sar plus ksar for graphing
- it can produce PDFs too.
http://ksar.atomique.net/index.html
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug




More information about the GLLUG mailing list