Message netf:121 -  Sent
Send-date: Tue,  3 Dec 1991 13:35:39 UTC+0100
From:       <Olav.Kvittem@delab.sintef.no>
To:         <netf@nic.nordu.net>
Message-ID: netf:121 2463*C=no;PRMD=uninett;O=sintef;OU=delab;S=Kvittem;G=Olav
Subject:    minutes CONS WG NETF3

Particpants :
	Matti Aarnio 	UTU/FUNET	<mea@funet.fi>
	Per Andersson 	CTH		<pa@cdg.chalmers.se>,
     	Mats Brunell 	SICS		<matsb@sarapis.sics.se>, 
	Johnny Eriksson SUNET/KTH	<bygg@sunet.se>,
     	Ole Frendved Hansen UNI-C	<ole.frendved.hansen@uni-c.dk>

Purpose:

Discuss the properties of X.25-networks and in particular IXI to create
an evaluation report on the usability of IXI.


Issues being discussed:

1. X.25 2 Mbps status.
Examples of switches/networks capable of running 2 Mbps X.25 given
(JANET, the WIN experiments).
Some figures from UNINETT showing that the delays in fact could
be improved dramatically by using X.25 fragmentation in a multihop network.
The speed of access-links from routers to switches was shown to be important to
the delay. UNINETT figures showed a decrease in 10% of single ttcp performance
using X.25-swtiches beween ciscos on 64 Kbps and 256 kbps backbones.

2. IXI status.
A brief summary was given on the toplogy

3. Evaluation of X.25(IXI) as backbone for IP  :

pros :
	X.25 is widely available and accepted standard
	large investment base
	congestion control
		window control on access line keeps the queues outside the
		network if proberly set.
cons (contrafavourable) :
	More boxes gives more management and less performance
	concatination of networks is bad for 
	 - flow control 
	 - management of parameter settings
	 - addressing 
		 - private network CAN'T have DNICs/PNICs (PTT restriction)
		 - no directory protocol for mapping of virtual addresses
	routing
	 - no protocols gives static routing configuration
	 - RPOA routing mechanism for DNICs standarddized but not implemented

      	costs - might be high for "small backbone" because PTT-type 
		switches are designed for a large number og attached DTE's not
		"networks" with a single line.
   
	implementation inconsistences between manufacturers
		revealed (but now solved) during IXI setup
		optional features variably implemented
		conformance requirements not the same in each country
	reliability
		reset/clear of circuit on X.25-link downs (no problem for IP)
	congestion control
	 - static window adherence obligation - not dynamically adjustable
	IP-routing
		when attaching n IP-routers to each other i creates
		n-1 circuits on each access link and n(n-1)2 circuits
		to watch in the network (for IXI it would initially be 
		20-30 and increasing) This will be trouble some to scale.
		So routers at central hubs must be used - causing traffic
		leave and REENTER the X.25-network (suboptimal)
	


A report will be (in retrospective : has been !) put together by Olav Kvittem.


Olav