>Access links are a somewhat different story:
>
>http://bianaoh.cc.columbia.edu/cgi-bin/cricket/grapher.cgi?target=%2Fmajor-uplinks%2Fnyser-uplinks%2Fpos0_0_0&ranges=d%3Aw&view=Bits
>
>(Upgrading the link would simply shift the plateau to whatever link
>speed we pick. There is no limit to the need for BearShare/Kazoo
>traffic.)
I'm not sure what this shows other than a saturated link. It does not show
to me that saturated edge links are a natural artifact of the Internet and
that edge wires will _immediately_ saturate at whatever speed they are
provisioned. Indeed I think it shows the opposite, that the network
services by an edge link does have a particular traffic volume requirement
and that a network's demand volumes are not infinitely elastic.
All this shows is that the growth trend in the traffic on this circuit hit
the 155M ceiling by November, and its time to consider multiple egress
links and balancing the traffic over them. Better choose a different
provider ofr the second edge link - that way you combine the immediate
benefit of greater external capacity with the risk mitigating factgor of
multiple external connectivity providers. But how to balance the traffic
over these multiple external links? Well there's always the technique of
announcing the same covering aggregate routes over both circuits and
announcing discrete sets of more specifics over each link. After all its
only the routing system that cops the consequent scale load, and nobody
_really_ cares about that - they'd rather head off into the desireability
of doing some form of QoS on backbones :-) :-)
Geoff
This archive was generated by hypermail 2b29 : Mon Aug 04 2003 - 04:10:03 EDT