According to Performance Tuning Guidelines for Large Deployments (HTTP) you should be able to calculate the number of http connections peak per second by taking a look at:

access_log files (The ones found at /opt/zimbra/log/ I suppose).

I suppose that it is a matter of a bash script with some of awk so that after taking out the day+hour I just leave the lines that have repeated hour+second string and just leave the ones that have more lines so that I get the peak by tunneling it to wc -l command.

Is the algorithm idea right?

Has anyone written this such a parser script? It would be nice if you could share it. Thank you.