[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference azur::mcc

Title:DECmcc user notes file. Does not replace IPMT.
Notice:Use IPMT for problems. Newsletter location in note 6187
Moderator:TAEC::BEROUD
Created:Mon Aug 21 1989
Last Modified:Wed Jun 04 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:6497
Total number of notes:27359

4798.0. "Network Overhead for Alarms/IP Poller?" by MAIL::CLAYTON (Merlin Clayton DTN 445-7217) Thu Apr 01 1993 11:30

Does anyone have a feel for the network overhead imposed by alarms rules
which poll entities for attribute information?  I'm interested in the size
of the packets (on average) for polling activities.

How does the Ip Poller function affect network overhead?  I assume that the IP
Poller just Pings all SNMP entities registered.  How would this differ from
a global alarm rule which looked at the ipReachability of every SNMP device?

I've also heard some discussion about autodiscovery consuming a great deal 
of network overhead.  Do we have any information to quantify this?

Thanks

Merlin


T.RTitleUserPersonal
Name
DateLines
4798.1IP Poller is faster...MOLAR::CHRISB::BRIENENNetwork Management Applications!Thu Apr 01 1993 14:3914
>  I assume that the IP Poller just Pings all SNMP entities registered.

  The IP Poller is actually mcc domain-based, so it'll poll all of the IP
  nodes in a particular domain (or domains).

>   How would this differ from
>   a global alarm rule which looked at the ipReachability of every SNMP device?

  The packet getting sent on he wire should be identical. The IP poller
  is just faster (we're talking something like an order of magnitude
  faster, and using less memory and cpu in the process).

  Overhead is negligible on a LAN. Polling lotsa nodes on the other
  end of a slow router is a different story...
4798.2MOLAR::YAHEY::BOSEThu Apr 01 1993 15:0722
>>How does the Ip Poller function affect network overhead?  I assume that the IP
>>Poller just Pings all SNMP entities registered.  How would this differ from
>>a global alarm rule which looked at the ipReachability of every SNMP device?

	The IP Poller sends out 50 ICMP Echo requests at a time and
	waits for the timeout and decides how many of the entities
	responded. Global wildcarded rules are serial on the other hand.
	Thus each entity is tested for reachability serially. 

	The size of the packets sent out by the IP Poller is 64 (data) +
	8 (icmp header) + 20 (IP header) bytes. Assuming no retries, and a 
	sufficiently large polling interval, the network traffic generated
	is :
	   (No. of entities) * 92 / (polling interval) bytes/sec

	So for 100 entities with a polling interval of 5 minutes the
	traffic generated is

		100 * 92 /300 = 30.67 bytes/sec

	Rahul.