Cisco has some guidelines on their web site at
http://www.cisco.com/warp/customer/701/38.html (You may need a CCO userID
to access the page). We poll the following MIBs from routers: avgBusy5,
freeMem, bufferFail and bufferNoMem. We poll and calculate the following
three metrics from every interface:
Utilization: half-duplex (ifInOctets + ifOutOctets) / ifSpeed * 800
full-duplex (MAX(ifInOctets + ifOutOctets)) / ifSpeed * 800
Capacity (discard rate): (ifInDiscards + ifOutDiscards) /
(ifInUcastPkts + ifInNUcastPkts + ifOutUcastPkts + ifOutNUcastPkts)
Quality (error rate): (ifInErrors + ifOutErrors) / (ifInUcastPkts +
ifInNUcastPkts + ifOutUcastPkts + ifOutNUcastPkts)
Depending on the type of interface (ATM, Ethernet, etc.) there are
additional MIBs you may wish to collect that are interface specific.
Thresholds: we use 75% for CPU utilization, 50% for interface utilization
(very conservative) and 1% for both capacity and quality. The threshold
values are related to the polling frequency. For example, if you poll every
five minutes, you may want a higher threshold than if you poll every fifteen
minutes.
To discover tendencies or trends in your network, you should store at least
six months of data, and generate some trend reports. This can quite a
challenge without some kind of purchased product to help you manage the data
and report on it. We have used INS' EnterprisePRO in the past which was
very easy to use and maintain. The reports were all accessible with a web
browser. We currently use SAS Institute's IT Service Vision product which
has a built-in interface to NetView-collected SNMP data. You can do trend
reporting "out of the box" or you can create your own custom reports which
requires some knowledge of SAS programming.
-----Original Message-----
From: Haakan Andersson [SMTP:Hakan.K.Andersson@ICA.SE]
Sent: Thursday, September 24, 1998 12:47
To: NV-L@UCSBVM.UCSB.EDU
Subject: How to be pro-active?
Hi,
I'm new to NetView AIX and snmp as well as networking. I've recently
taken
over
the responsibility for NetView AIX (running v 4.1.2 with NWays
Manager,
will upgrade
soon to 5.0).
We are currently migrating to Cisco routers in our WAN.
At this point, the operators/helpdesk can almost only see if a node
is up
or down.
Very little configuration and adapt-ing (is that the right word?)
have been
done yet.
The issues I will start to deal with now is;
- how can we in NetView discover network problems before they become
alarming,
how can we be pro-active?
- for instance, when setting thresholds of the traffic through
routers,
what threshold
values is propriate?
- what "parameters" are important to monitor in a router?
- what MIB variables are of interest concerning the topics mentioned
above?
- how to discover divergent tendencies in the network
I guess these topics are more snmp-, "technical networking"- and
router
admin/managing
issues. However does anyone know of any good books that discuss
these kind
of
topics? Maybe someone has some tips and hints regarding this?
Best Regards,
Hakan Andersson
ICA Handlarnas AB, Sweden
|