craigis Posted August 18, 2011 Share Posted August 18, 2011 Hey guys, I was just wondering how your system calculates the server loads... By this I mean what sort of things are taken into consideration when the number is calculated... As I'm writing this Stevies load is 156.78 but that's due to the upgrade. Anyway I was just curious as to how this is done. Thanks Quote Link to comment Share on other sites More sharing options...
Piotr GRD Posted August 18, 2011 Share Posted August 18, 2011 So called "server load" is the average number of jobs (processes) in the run queue (waiting for CPU share) or waiting for disk I/O (load/write the data). So we can simplify it and describe as CPU and I/O utilization on the server. Inside of cPanel you'll find the average server load for last 1 minute, the tool displayed as the images on the forums uses the value average for last 15 minutes, if I'm right. And on provided by me server monitor you can find graphic presentation of three values - average for last 1, 5 and 15 minutes, so all that are reported by the server itself (well, 2 servers in here). Quote Link to comment Share on other sites More sharing options...
jje Posted August 18, 2011 Share Posted August 18, 2011 I believe the server load is read from the third number listed in /proc/loadavg. The same server load can also be retrieved by admins with the 'uptime' command. @Piotr - Actually, I think the server load we post is the current up to date server load. However, this might soon be changed to 15 minute to reduce the time taken to retrieve the load; I think xaav might be working on this now not sure though. Quote Link to comment Share on other sites More sharing options...
Guest xaav Posted August 18, 2011 Share Posted August 18, 2011 @Piotr - Actually, I think the server load we post is the current up to date server load. However, this might soon be changed to 15 minute to reduce the time taken to retrieve the load; I think xaav might be working on this now not sure though. Already done. The load displayed here is cached for 5 minutes. Quote Link to comment Share on other sites More sharing options...
Piotr GRD Posted August 18, 2011 Share Posted August 18, 2011 (...) number listed in /proc/loadavg. (...) And this is exactly what I am describing - these numbers are the average numbers of processes waiting either for the CPU share or to access the disk to load or write some data. Third number mentioned by you represents the average number of processes for the last 15 minutes. While the images on the forums won't be reloaded every minute by most of the people I think that displaying average for longer time is a good choice made by the person that created it. Seeing a temporary 1 minute peak while 1 minute before and 1 minute later server load is low and fine could be misleading. On the other hand on my graphs you can view the 24 hours history, so viewing such short 1 minute peaks can be sometimes useful. Some can be missed, though, as I query the servers every 2 minutes, but anything significant will be reported by the numbers of average load for longer time periods (5 and 15 minutes). edit after seeing xaav's reply: ... I think that displaying average for longer time is a good choice made by the person that created it, especially if the retrieved from server value is cached. Quote Link to comment Share on other sites More sharing options...
craigis Posted August 20, 2011 Author Share Posted August 20, 2011 Ah ok, I wondered if it was a system generated value or if you had to produce the system completely... but its still interesting to know Anyway thanks for the quick replies Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.