What about our server performance metrics? Would it be enough in case of board users increasing after finalizing real score cluster? We’ve prepared few CPU usage report graphs:
As you see 100 user sessions loaded CPU by the only 5%. We constantly working on script and DB optimization. That’s why CQ WW CW contest 100 users sessions got CPU load 1.5 times less than ARRL SS CW with just 60 user sessions.
Anyway there is a lot of room for improvement! Take a look at the hourly CPU load graph. There are some CPU load spikes produced by a script running in a beginning of every hour calculating the “Highest rate” for a contest. It does a search over 100K log entries withing few seconds:
So are we ready for the cluster? The answer is “Yes”. The biggest number of online score board users I’ve ever seen was 600 during the last CQ WW CW on the cqcontest.net. If we do some calculations we could figure out the COSB CPU load would be around 30-35% with 600 online users. Anyway there is a lot of room for improvement! The current server is a modest VPS. So RAM and vCPUs could be easily upgraded. Moreover we designed our server side scripts such a way to be ready to split incoming traffic to a front servers cluster. More to come …