Hi all,
I am gathering a lot of information from many servers and I want to create charts that will easily show what goes wrong and when.
Here is the information I can have access to :
- CPU usage in percentage, measured every 30 seconds
- Number of logged on users, measured every 30 seconds
- Per process CPU usage (top 5 averages for the day), measured every 10 seconds
- Per process CPU usage (top 5 consumption peaks for the day), measured every 10 seconds
- Memory usage in percentage, measured every 10 seconds
- Network usage in bytes / second, measured every 10 seconds
- Disk usage in bytes / second, measured every 10 seconds
- CPU usage in percentage, measured every second
Naturally, data measured every 10 seconds can also be computed for 30 secs if needed
Right now, I create graphs that show CPU usage and logged on users, plus historic averages and data from the day before :
Excel-chart-ex1.jpg
But since I just added tons of extra info, I need to rethink this.
I want to be able to see if there is a big change in the pattern at first glance and once a problem is detected, to easily see what is wrong (for example, detect a high CPU usage, which would lead to check the top processes, to maybe realize disk usage is going too high...)
So I thought about combining data on some more graphs (I would probably keep the existing ones).
One thing I thought of would be to use an area graph showing RAM usage as background for a line graph showing CPU, Network, and Disk usage. I would use the same method for top processes.
But this seems quite hard for 2 reasons :
1. I can't find a way to mix area and line charts
2. I am stuck with only 2 Y axises :D The numbers being on very different scales, some results would be barely visible.
Any suggestion would be greatly appreciated.
Thanks.
Bookmarks