Ultra1 (fwd)

[Robert Dinse <nanook@eskimo.com>]

>     Ultra1 has run out of memory twice this morning.  Something is causing
>traffic to be way above normal and I have not yet been able to figure out
>what.

     I found one users script was stuck in a loop trashing the server pretty
bad, but even with that fixed there is still heavier than normal traffic and I
don't know why.

     Both mail list and web traffic are heavy and I don't see any common
denominator.

     The reason heavier than normal traffic causes problems is that when
requests for web pages, etc, come in at a rate that is higher than the web
pages can be served out because the network links are saturated, it causes data
to be buffered, and while momentary traffic spikes and buffering of that data
is normal, if traffic is sustained at rates higher than the network connection
speed it will eventually run the machine completely out of memory which is what
is happening this morning. 

     The machine already has 640MB of memory in it, so it's not like it's
starved, the real problem is inadequate bandwidth.

     Normally at the beginning of a new year we see higher traffic, this has
always been the case, but usually by mid-March it dies down to more normal
levels.  This did not happen this year, that is traffic increased and stayed
heavy.

     I've ordered another T1 to address this but am waiting on Sprint/Qwest to
install it (Sprint provides the backbone, Qwest the physical access link to
Sprints POP).