--- Rich Adamson <[EMAIL PROTECTED]> wrote: > > do have anybody experience with a mrtg on a windows system which measure > > about 200 OID at the same time? > > How many power (cpu, ram and so on) should have this windows system? > > Any hints to do this? > > > > In the past a measured about max. 50 different values on one PC wiht > > windows and mrtg. > > If all 200 OID's are in the same polling app, the cpu isn't going to > make much difference. The issue will be the round-trip time required > to poll a remote device and obtain a response. The polling happens > sequentially, one oid/device after another. > > If the polling cycle is every five minutes and all oid's/devices are > within the same polling app, and "each" oid requires a 100 millisecond > poll-response time, then the best one can do is something around > 25,000 oid's (assuming I didn't screw up the math).
I think your math might be off by a bit... 100ms poll-response means you can poll 10 OIDs per second; a five minute cycle is 300 seconds, which means you max out at about 3,000 OIDs per box if you're doing them all linearly. I've found that I can run about 4 groups of devices in parallel on a single box before the disk I/O becomes a bottleneck, which gives about 12,000 total OIDs per 5 minute cycle; each interface needs 2 OIDs, plus every router gets its uptime info polled, so you're well below 5,000 interfaces on a reasonable system every 5 minutes. > The real key is knowing exactly what your poll-response times are, > and what else is running on that same box that consumes processor > cycles. Actually, I've found that disk I/O is my main bottleneck. The actual work of sending and receiving the queries is relatively trivial compared to updating the on-disk records; obviously, using RRD instead of log format makes a large difference here. Matt -- Unsubscribe mailto:[EMAIL PROTECTED] Archive http://www.ee.ethz.ch/~slist/mrtg FAQ http://faq.mrtg.org Homepage http://www.mrtg.org WebAdmin http://www.ee.ethz.ch/~slist/lsg2.cgi
