nanog mailing list archives

Re: Measure overall network availability


From: Suresh Ramasubramanian <ops.lists () gmail com>
Date: Fri, 7 Jan 2005 12:09:10 +0530


On Fri, 7 Jan 2005 12:23:48 +0800 (CST), Joe Shen <joe_hznm () yahoo com sg> wrote:

is there any recommended method to measure overall
network availability?


Customer complaints that they cant reach you

Network availablity from where?  Maybe a script that polls visiblity
of your AS from various looking glasses, and also sends you ping /
traceroute times from these looking glasses, on a periodic basis.

Not too periodic unless you've talked to the LG operator and arranged
that they know about the test bot you've written to automatically
query their looking glass on a regular basis.

Currently we use packet loss rate as indication of
network availability, but to my understanding this
just means the possiblity of e2e communication degrade
but not the network availability.

Maybe maintain a few 1U colo boxes (cheap!) in data centers on
selected networks around the world, from where you want to measure
reachablity .. run nothing except nagios or some other monitoring app
for measuring availablity of services like http, smtp, etc that you
want to know are available or not,

All these nagios reports can be configured to land in a central
monitoring console in your NOC.
 
-- 
Suresh Ramasubramanian (ops.lists () gmail com)


Current thread: