|
|
|
|
|
|
|
|
|
|
|
|
| home | help | back | first | fref | pref | prev | next | nref | lref | last | post |
Date: Fri, 3 Jan 92 21:13:49 EST From: martillo@azea.clearpoint.com (Joachim Martillo @ azea) To: rma@tsar.cc.rochester.edu Cc: farber@central.cis.upenn.edu, com-priv@psi.com, In-Reply-To: Richard Mandelbaum's message of Fri, 03 Jan 92 11:09:08 -0500 <9201031609.AA00299@tsar.cc.rochester.edu> Date: Fri, 03 Jan 92 11:09:08 -0500 From: Richard Mandelbaum <rma@tsar.cc.rochester.edu> I never cease to be amazed by assertions of how easy it is to manage production level operations over the US using new technology which is unproven!!! I was not asserting that management of production level operations over the US with new technology is easily. I was merely expressing surprise that anyone considered the development of a DS3 router a significant feat. If I started with a multiport router which supported 8 or more ethernet or token ring interfaces and therefore probably had sufficient internal bandwidth for routing from a DS3 interface. I could see about a month for design of a DS3 card, a month for layout/CAD, about 3 weeks for prototypes, a month or two for bringing up the software and debugging the board (using existing serial interface software modified for the unique characteristics of DS3 line interface if any), a month alpha and then two months beta. Thus I would estimate about 8 months to product starting from the base of an existent multiport router. I will grant that bringing up a nationwide DS3 backbone using DS3 routers presents some problems, but I was not addressing that issue. Also since there is but one DS3 backbone network as of yet, we have no standard for comparison that alternate design choices might have lead to a more flexible, less costly and more easily manageable backbone network. It is simply too early for congratulations and I think we are seeing the ATT/IBM performance measurement syndrone which is the inability to determine whether the management should be congratulated or fired for achieving X% revenue growth because there is no way to determine by way of comparison with similar corporations that different managements would have done better or worse. ____________________ Date: Wed, 1 Jan 92 18:27:25 EST From: farber@central.cis.upenn.edu (David J. Farber) Posted-Date: Wed, 1 Jan 92 18:27:25 EST Just to bring a bit of facts to some of the discussions. When the DS3 network started , DS 3 routers were not easy to build -- they still are hard. IBM did a fine job of doing that!! We can mayb e do better now but that's now!! Maybe, I don't understand all the intricacies, but I am puzzled at the assertion that DS3 routers are hard to build. DS3 is a 45 megabit service. Several vendors have been selling for several years ethernet/token ring/fractional T1 routers which can handle aggregate bandwidths of approximately 100 Mbs. Well, taking 4lan interfaces off of one of these boxes and installing a reasonable serial interface which can handle T3 speeds should produce a quite good DS3 router. There is no obvious reason why the government had to provide IBM, which is not a router company, with its own private test bed for router development at government expense. An open bid would have made more sense. Dave (one of the authors of the NRC report) Joachim Carlo Santos Martillo Ajami
|
|
|
|
|
|
|
|
|
|
|
|
| home | help | back | first | fref | pref | prev | next | nref | lref | last | post |