I am looking for a way to log server performance data to be viewed later as a graph. This will help we tell the times which our servers are under the most load as well as show me possible performance bottlenecks. I would like to monitor things like cpu load, memory use, nework activity and disk activity. The operating systems I need this to run on are Windows 2000 Server and Windows Server 2003. Can this be done with the Performance Monitor snap in that comes with windows? If so does anyone know of a good guide because I can't figure out how to use it?

Thanks in advance

I usually monitor things this way. Just need MRTG and some scripts. You can monitor anything you want.

That looks promising. Does it keep a history so that I can look at the statistics for a previous day? Or can I look at the data for a whole week. For example it would be nice to look at the processor load and memory load in a graph for the whole week so I can see if some days the load is more than others and track usage patterns?

Yes it keeps a "history". Just look at the linked page for the very first image here. There's even a yearly graph on there.

It's meant to monitor SNMP devices, but you can make it graph really anything you want using simple scripts (gotta understand basic scripting though - which I would hope you do if you're monitoring servers).

Also, check out RRDtool for somewhat nicer graphing.

There's a bit of work involved to make it all work, but it's totally worth it. You can find tutorials, guides, FAQs and all kinds of resources using good ole google.

This looks really great. I can't believe I have not looked into this sooner.

Thanks a lot. This will be a huge help to me.

You're welcome.

Yeah, it's quite nice. I mean, one could do it all by hand, it's really not THAT hard to make something simple that would work. Have scripts that collect info (SNMP would be a pain though) and dump it into logs, then log parsers, math and some GDI+ stuff for graphing, etc. But then one has huge logfiles taking lots of disk space and that take a while to process. So you'd have to develop some round robin database (RRD) equivalent or such. Optimize, maintain, support, document and bugfix it all. Add new features. Add some scripting or config engine so you can get it to do fancier stuff (especially for the graphing)... An awful lot of work for what seems like such a simple task. And here all we have to do is get it to collect the data, that's about it.