A few observations about looking at TOP that I hear people asking are regarding what does the % mean, and why can I have over 100% when looking at CPU usage.

Well, its sort of simple. Load averages are the average of the load number for a process in a given period of time. A load number is merely a metric that is an average that takes into account CPU usage (or consumed time-slices) by the process. These could be:

• Operational (active) processes running on the CPU
• Runnable, but waiting (could be waiting on a spin-lock
• Sleeping (e.g., waiting/blocked on IO)

I won't delve into how the arithmetic of how the values are calculated, but there is an important note here: CPU usage can be over 100%. What!? The idea is (and its also due to the age of the tool) that the value is divisible by the number of cores you may have. If the usage is 100% and you have two cores, only 50% of the total overall CPU time was used (or 50% was free).

Other notes are, make sure to analyse trends in utilisation. Short term "bursts" can be for a variety of reasons such as the spawning of a process or a bulk operation, but increasing trends should be more worrisome.

When looking at a specific process's memory consumption:

I tend to look at these variables:

For more information about these variables, please see this excellent answer on Quora