> Hi, our ISP has been using MRTG for a while (still at 2.13.2), great > stuff, thank you! Our userbase counts a little over 1200, that's 1200 > HTML pages and 4800 PNG files to take care of. The problem is that the > pretty powerfull Dell PowerEdge (dual Xeon) is taking almost > 50 seconds of full CPU time to finish drawing the graphs every 5 minutes.
The way to speed it up is to move to RRDTool. This makes a huge, huge speed gain. You still use MRTG to collect the data, it is just stored in the .rrd files instead of in .log files (and you can convert the .log files to .rrd to preserve your history). Of course you then need a cgi frontend such as routers2.cgi (or mrtg-rrd, or 14all) to display the data ,and the .png files are only generated when required. Further speed gains can then be obtained by using mod_perl or speedycgi. You also get extra reporting capability when using RRD files. You can get more performance by getting very fast disks. MRTG bottlenecks on disk IO, so the fastest disk and controller you can get is advised. Also, if you are using UNIX, then mount the data filesystem with the noatime option, add extra directory indexing, and use subdirs for your .rrd files to save MRTG from having to search big directory indexes. Adding more memory can help to a point (due to directory caching and being able to hold multiple threads of perl at once) but disk IO is the key. We have a dual 2.5GHz Xeon server that can cope with (currently) 6885 targets, and still have capacity for more. We have a fast Adaptec controller with some speedy external SCSI disks (sorry, cant remember the spec offhand) Steve -- Steve Shipway ITSS, University of Auckland (09) 3737 599 x 86487 [EMAIL PROTECTED] -- Unsubscribe mailto:[EMAIL PROTECTED] Archive http://lists.ee.ethz.ch/mrtg FAQ http://faq.mrtg.org Homepage http://www.mrtg.org WebAdmin http://lists.ee.ethz.ch/lsg2.cgi
