This is discussed in more detail in the MRTG book here http://www.steveshipway.org/book (sorry for the advertising)
Basically, it depends on how many Targets you have, and whether or not you use external plugins or just SNMP checks. So, if your 100 machines are being checks for 1 network card and 1 cpu stat, that would be 200 Targets, probably using SNMP only. The main bottleneck for you will be Disk IO. You should get the fastest disks you can get as slow disk IO slows down polling and means more threads are running, increasing memory requirements. A 200-Target installation is relatively small. We run about 14,000 Targets on our dual-Xeon 3GHz linux server with 4GB RAM, with still some spare capacity. You could certainly manage with a much smaller server. If running under windows you have a slightly higher requirement but not much. Note that, for performance, you should use MRTG with RRDtool and a frontend such as Routers2, 14all or mrtg-rrd. While you can run in native mode, performance is far worse and you will find that you need a much larger server to cope with the load, as in Native mode MRTG will recreate all the graph images every 5 minutes, rather than only on demand. A native-mode MRTG may need 10 times as many resources as an RRD-mode installation. Hope this helps, Steve ________________________________ From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Huh, Jin (FXKIS.KOR) Now we are trying to implement MRTG server and monitor around 100 machines of network traffic usage and CPU performance. (Servers and Network devices) Please let me know appropreate hardware requirement for MRTG server.
_______________________________________________ mrtg mailing list [email protected] https://lists.oetiker.ch/cgi-bin/listinfo/mrtg
