[c-nsp] High SNMP CPU with SXH. Is SXI any better?

[Karl Gaissmaier]

Hi all,

Phil Mayers schrieb:
>> I haven't looked into it, but perhaps you can find a cisco specific
>> mib, maybe cef or mls specific that doesn't have this performance
> 
> CISCO-SWITCH-ENGINE-MIB::cseCef* are the ones. However, they contain 
> basically the entire FIB & adjacency entries. They're faster to walk, 
> but their size can make the whole process slower :o(

If you use the cseCefAdjacencyTable, the number of entries are
similar to the ipNetToMediaTable.

My tests showed, that fetching the ipNetToMediaTable
stresses the route processor at ~90% cpu load.

Fetching the cseCefAdjacencyTable stresses the switch processor
and the route processor together, but now both processors at
~50% cpu load.

> 
>> penalty?  Otherwise, I bet a query via clogin outperforms the snmp
>> table.
> 
> It certainly does. ~10 seconds versus ~2min 30sec on our busiest router.

Example:

a column from the MIB-II ipNetToMediaTable
==========================================

unix$ time snmpbulkwalk -v 2c cat65-box 1.3.6.1.2.1.4.22.1.2

5595 entries

real	0m11.699s
user	0m0.128s
sys	0m0.020s

a column from the CISCO-SWITCH-ENGINE cseCefAdjacencyTable
==========================================================

unix$ snmpbulkwalk -v 2c cat65-box 1.3.6.1.4.1.9.9.97.1.8.3.1.4

6044 entries

real	0m4.055s
user	0m0.152s
sys	0m0.028s

> Cisco - Offering the very best in 1970s Network Management Technology...

ACK!

Best Regards
	Charly
-- 
Karl Gaissmaier
Kommunikations und Informationszentrum kiz
der Universität Ulm
Abteilung Infrastruktur
SG Netzwerk und Telekommunikation
89069 Ulm
Tel.: 49(0)731/50-22499 Fax : 49(0)731/50-1222499