[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference 7.286::fddi

Title:	FDDI - The Next Generation

Moderator:	NETCAD::STEFANI

Created:	Thu Apr 27 1989
Last Modified:	Thu Jun 05 1997
Last Successful Update:	Fri Jun 06 1997
Number of topics:	2259
Total number of notes:	8590

290.0. "Design query for the panel." by YUPPY::MCINTYRE () Tue Jun 25 1991 06:56

    
    
    I work in DECsite London and have a configuration question for the FDDI
    forum.
    
    A customer of mine has asked for a design incorporating four dual
    redundant 100MB rings.
    
    Two rings (A and B) will run round the main campus as a backbone for
    20 odd buildings with concentrators etc. providing access where required.
    
    It is assumed at this time that the campus will be split between the
    two rings i.e. some building fed from ring A, some from ring B.
    This is a redundancy and a security feature.
    
    The other two rings (C and D) will run round the inside of one of the
    buildings (computer centre and support centre), one ring providing a 
    backbone for the VAX systems and RISC machines, the other used to 
    interconnect to 10MB devices and SMART Hubs for user access.
    Once again we are assuming that security,redundancy and load sharing
    are the customers requirements.
    
    The computer centre building is on the same campus as the rest of the
    buildings concerned and separated by a distance of 300 metres or so.
    
    Can we link all the rings together without introducing bottlenecks ?
    
    It would be best if we can provide dual redundant links between the
    computer centre rings and the rings running round the campus.
    
    The design has not been set in concrete at this time so if it isn't
    going to work as described above it may not cause too much trouble..we
    will redesign the whole thing for the customer and charge day rate for
    doing it.
    
    PS Is IEEE802.6 capable of FDDI to FDDI interconnection ??
    
    Any assistance would be most welcome.
    
    Firstly let me say the design has not been set in concrete at this time
    and may change..but assumin

T.R	Title	User	Personal Name	Date	Lines
290.1	does this make sense ?	VLAB::CARR		`Tue Jun 25 1991 15:18`	26
	re .0 >> It is assumed at this time that the campus will be split between the >> two rings i.e. some building fed from ring A, some from ring B. >> This is a redundancy and a security feature. How does this provide redundancy ? If some sites are connected to ring A and some are connected to B, then a failure on either ring could leave those sites out in the cold. Having another ring around that doesn't connect to those sites doesn't seem to provide redundancy at the ring level anyway. >> Can we link all the rings together without introducing bottlenecks ? How are you planning to link the rings together ? I'm not a network guru, so maybe I'm missing something here, but having 4 FDDI rings doesn't seem to support the stated goals of redundancy and load sharing. Maybe someone with more experience can shed some light ? Regards, Denise
290.2	reset......	YUPPY::MCINTYRE		`Wed Jun 26 1991 11:03`	15
	The problem is the customer wants to load share amongst 4 FDDI rings..what I want to know is can we do this without using 10/100 bridges back to back between the various rings ?? And if we do link the rings together with 10/100's is it a legal configuration ? PS 802.6 and Cell Relay are technoligies that have been banded about as capable of linking 100MB rings AT 100MB speeds ....anyone have anymore info on these methods ??? Or is it all vapourware at this time ??
290.3	NO ONE ?????	YUPPY::MCINTYRE		`Wed Jul 03 1991 04:14`	9
	Does anyone have any ideas on this problem ??????? Regards Tom
290.4	it works anyway	DELNI::GOLDSTEIN	Networks designed while-u-wait	`Wed Jul 03 1991 09:53`	33
	Somebody please correct me if I'm wrong, but if we're dealing with 802.1-compliant bridges, and I think that's the kind we make (no smiley needed), then there is no load-splitting available. The 802.1 spanning tree determines exactly one path between any two LANs. If that path breaks, it finds another. Parallel links are generally "hot standby". In traffic engineering, you always have bottlenecks. I characterize them as "dumbbell" and "football" (this only works for the American kind!) networks. Dumbbell networks are wide at the outside and skinny on the inside; football networks are the opposite. The common LAN-WAN-LAN arrangement is a dumbbell, since WAN bandwidth is costly. Ethernet-FDDI-Ethernet is a football. In a dumbbell, congestion occurs getting on to the link. In a football, it occurs getting off. You've moved the problem. But it can't be avoided. Fortunately, our protocols are designed to support this. DECnet, IP and OSI all adapt to the currently-available bandwidth. A bottleneck anywhere will have the same effect. 802.6 is just another inter-LAN technology. It runs at many speeds, unlike FDDI. So you can run it dumbbell, football or "matched", but the latter just means that you are as likely to congest at the access as at the egress. ATM cell relay (the B-ISDN kind) isn't here yet, but it adds a new wrinkle: If you use the official protocols, then you'll have a risk of losing cells, and each lost cell causes the entire packet to be lost. The onset of congestion collapse as a result of this is left as an exercise to the reader. Hint: Discard unit is smaller than retransmission unit. So if the customer pitches alternatives to FDDI, it isn't solving the problem. If it's a problem at all. Unless you know how much demand you'll really have for bandwidth at any give subnet, you can't tell if it'll be adequate.
290.5	NIS 600???	YUPPY::MCINTYRE		`Thu Jul 11 1991 08:45`	8
	ok..another question... Will the Hastings box (NIS 600) be able to link 100MB rings ? Tom..
290.6	As a Brouter	MARVIN::DAVISON	Eric Davison	`Thu Jul 11 1991 08:49`	8
	We are building a FDDI card for the DECnis 600, this will have Routing and Bridging capabilities. You'll be able to put at least 2 into the box, so you will be able to broute between two FDDI LANs. Availability isn't announced yet, contact Paul Keating or Richard Benwell (both on JANUS::) if you need the date. Eric
290.7	@ 100MB??	YUPPY::MCINTYRE		`Thu Jul 11 1991 11:10`	9
	thanks for the info on the nis...do you know if the brouting will be at full fddi speed ??? regards tom
290.8	Send all questions to Product Management please	JANUS::BARKER	Jeremy Barker - T&N/CBN Diag. Eng. - Reading, UK	`Thu Jul 11 1991 11:33`	6
	Re: .7 Try asking the people named in .6. They may be able to give you an answer. jb
290.9	maybe not quite 450000 pps on transmit :(	MARVIN::DAVISON	Eric Davison	`Fri Jul 12 1991 08:32`	12
	As Jeremy points out, detailed questions should go to Product Management (or whatever level of ANC/RNC/whatever they're called nowadays). Hopefully the announcement package contained some of this info, presumably (it had better be) there will be information on DECnis 600 at the next Net U. However I'm fairly certain [big :)] that we won't be forwarding at the 450,000 packets per second that we can receive (ie filter) at. We are aiming to saturate the datalink for reasonable packet sizes. Eric