Super User is a question and answer site for computer enthusiasts and power users. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I'm doing some reading on SNMP and have Stallings' book "SNMP, SNMPv2 and RMON" (1996).

He shows how by using some basic math you are able to "determine the maximum number of stations that the management station can handle when engaged in full-time polling" (p. 194, 7.5.4 "Polling Frequency").

He gives the following:

N <= T/V

N = No. of agents
T = desired polling interval (in seconds)
V = average time required to perform a single poll

He then gives a worked example:

"The example consists of a single LAN, where each managed device is to be polled every 15 minutes. Assuming processing times on the order of 50ms, and a network delay of about 1ms (packet size of 1,000 bytes, no significant network congestion), then V is approx. 0.202 sec. Then:

N <= (15*60) / 0.202 = approx. 4,500."

I find this quite unclear. How do you get 0.202 sec from the information above?

Are there any other ways to determine how many stations can be handled by the management station?

Regards, ILMA.

share|improve this question
I could see .102 seconds... 1ms for the packet to get to the polled station, 50ms processing time, 1ms for the return packet, 50ms processing time on on the management station. Unless somewhere before the example he states that management stations take about 150ms to process the returned data. – cpt_fink Apr 7 '13 at 3:53
up vote 0 down vote accepted

Processing time to prepare the outgoing snmp query (50 ms), communication (1 ms), processing received SNMP packet (50 ms) = .101 s, which is 1/2 the transaction. SNMP processing ought not to take precedence over more fundamental operations, such as packet forwarding. As such, the responses are asynchronous. However, even though the numbers used are not the clearest, the formula is sound. In my experience, calculating the average response time would be a study unto itself. The 1 ms delay essentially means all hosts are on a simple switched LAN, with no routers, wi-fi, or long haul ethernet over fibre. In practice, one polling station is set up, and the intervals, timeouts, and verbosity of the queries are adjusted until no artificial alarms are generated by missed polls, and then depending on the business requirements, either the poll frequency or the poll complexity can be increased to target levels. Polling a 48 port switch incurs a much different management overhead than polling a small router, or a cable modem, so be prepared to monitor snmp packets in and out of the polling station to see what your individual situation is, and adjust parameters and distribute polling duties to meet your needs. I guess it would also be safe to say that 4500 hosts/15 min is an optimistic upper boundary.

share|improve this answer
Thanks, Nevin W. – I_lost_my_last_account Apr 9 '13 at 14:19

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .