 The magazine of the Melbourne PC User Group Internet Service Newsletter George Skarbek gskarbek@melbpc.org.au |
 |
Automated Notification of Poor Performance
During a period of poor performance Bob Thomas asked in melbpc.general newsgroup: "Is it possible to have some sort of automated process that keeps an ongoing record of the average data flow?"
This is a good question and a valid request. Rather than waiting for a message on the MOTD (Message Of The Day) from the Internet admin team informing our users that a problem exists, you can look at it yourself. Jason Lee set up this facility over a year ago and it is accessible from the MOTD page under System Statistics and then Cisco router statistics.
Members of the admin team do not have automated warning messages sent to them. We generally look at these graphs first and some other more complex parameters that are only accessible to users with root access to the Unix computers.
When you bring tip this page there appears to be a lot of data. However the
two most important graphs to view arc the second and third, namely:
cuscus.melbpc.org.au (connection to eisa): Serial 1/2
cuscus.melbpc.org.au (mpc3143-I.gw.connect.com.au): Serial 1/1
|
|
|
|
If the dataflow on the first one mentioned should drop to below say, approximately 800, and the second one mentioned has not picked up a substantial part of the load, then we have a problem. However, this could be at our end or the other end or even further downstream, but a problem exists and the data flow is curtailed for some reason. Be aware that these rates will change with time and if we sign up with a new wholesale supplier in the future, this will change the pattern dramatically.
If you click on the blue description or graph you will obtain graphs for daily, weekly, monthly and yearly summaries.
Note that the corollary is not true. Just because these is a lot of data flowing this does not always guarantee a fully healthy system, but this is a very good start as far as automated reporting goes. Our Internet service runs on seven main computers and there may be problems with some services such as mail or news that carrot be diagnosed just from these graphs.
However, this statistics page is the best automated system that we can provide to let users know of possible problems.
As to the actual cause of poor performance in mid December 99, it was caused by a change to another ISP'S router (not ours) in a city building that transferred all the news via our EISA link and not via the spare capacity we have on our 2 Mbit link into Collins St. As the volume of news is very high, this resulted in the "curling" traffic not being able to get through.
Another set of statistics that may be of interest to subscribers is to see when the phone lines are busy. From the MOTD, System Statistics, click on Melbpc Dial-Up Server statistics. These graphs are easy to understand. They show how many phone lines are in use for each of our phone numbers.
Reprinted from the February 2000 issue of PC Update, the magazine of Melbourne PC User Group, Australia
