Re: mbone/NSFnet migration (fwd)

[Curtis Villamizar <curtis () ans net>]

In message <199410201452.AA04207 () zed isi edu>, bmanning () ISI EDU writes:
> > I would rather see a discussion of the estimated load at the NAPs.
>
> Rat-hole !!!  (of course this is a tasty topic all by itself. I think
> there will be some discussion of this in Curtis and Johns presentations)
> I am uncomfortable even w/ 100Mbps broadcast media exchanges.  We might be
> safe w/ 155Mbps P/P  exchanges.

That's John's rathole, not mine.  :-)

> > Consider the worst possible migration sequence where (I'm guessing)
> > BARRNet, NC-REN, NEARnet and Sesquinet transition first.  What will be the
> > forecast load at the NAPs?  Is the calculation quick enough to run
> > for all 2^9 splits of the MCInet customers? (Assume Sprintlink customers
> > and other connectivity havn't moved yet)
>
> I think that MCI is collecting data on this now.

Really?  Based on the net-net matrix?  Thats an average over 30 day
periods.  How is this converted to peak hour load projections?  We
have SNMP link utilizations in 15 minute increments that could be
averaged over an hour to get a rough conversion factor.  We could
probably provide access to this data (Merit should already have access
to it, I think) or if not provide the conversion factor if asked.  In
any case, this is John's topic, but I'm quite interested in both the
bottom line numbers and the method (so we have a better idea how much
to trust the bottom line).

Curtis
- - - - - - - - - - - - - - - - -