Just wanted to make sure I understand Joe’s statement about CPU load monitor he made here:
Joe says: For example, on a heavily loaded server, with many processes competing for CPU load, the load average numbers can go above 1.0. Well above 1.0 in some case--a heavily loaded Squid proxy server can hit 5 or 6 for hours at a time. This doesn't mean that the machine is using 500% of the CPU, it just means that there are five processes that want 100% of the available CPU
I was confused because the number on the system information page in virtualmin shows load in terms of 1.0 or less (sometimes more if high load). However, the new svg graph (which looks great) shows numbers in terms of 100 +/-.
I understand the concept that the number can appear higher than 1/100 if there is high load, but on the graph page I can see spikes as high as 300 or more when the sys info page would show something like 1.7.
Geoffrey<br><br>Post edited by: djhomeless, at: 2007/09/17 06:44
It’s the same number and the same concepts apply–we just multiply it by 100 to get it into integers (the timeplot graphing library that we’re using is currently unable to deal with decimals). It probably ought to max at 100%, since there can’t possibly be 300% usage.
Again, it just means you have some number of processes that would use 300% of the CPU if that much performance were available. You can’t pay too much attention to it, because load monitoring on modern CPUs isn’t getting any easier or more clear, but if you’re consistently seeing high usage, and performance feels sluggish for any service, then it’s probably worth figuring out where the CPU is going and fixing it.
The default Virtualmin configuration can be quite demanding of CPU for mail processing, for example. We’ve begun to adjust our thinking on that a bit, as more of our customers are running higher load systems with hundreds rather than dozens of virtual servers and with thousands rather than hundreds of mailboxes. But, it only makes sense to tweak with measurement. Dig deeper using the Running Processes module or, better, running “top” from the command line. Find out where the CPU (and memory) is going, and we can help you tweak your settings to make things run more smoothly (if things are actually causing problems–load is, again, a slippy thing, and high load every once in a while is normal and not usually cause for concern).