Load average is the standard unix load metric. Its the average number of runnable jobs in the scheduler's run queue taken over a specified time period. This can indeed go above 1 per cpu. It is different than "the number of cycles used for running processes". There is a "load" figure on the summary page which is the load / #cpus.
-Federico ----- Original Message ----- From: Karl Kopper To: ganglia-developers@lists.sourceforge.net Sent: Thursday, February 12, 2004 3:01 PM Subject: [Ganglia-developers] Load average and Ganglia. Hi Ganglia Developers, I'm working on a book about clustering, and one chapter is about Ganglia. I notice in the documentation for the Ganglia Web Frontend that the load average is taken to be the same as CPU load or CPU utilization. The color coding scheme for the cluster nodes that go "red" when the load average goes above 1 (when divided by the number of CPUs) does provide an indication of a more heavily loaded node, but it does not mean the same thing as 100%+ utilization as the documentation states. Right? You can't have CPU utilization > 100% that doesn't make sense. Did I miss something? --Karl
