[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

637.0. "Full Duplex from FDDI Newsgroup - any comment" by VIVIAN::MILTON (Invisible person it seems!) Wed Jul 08 1992 09:18

Article 432 of comp.dcom.lans.fddi:
Newsgroups: comp.dcom.lans.fddi
Path: uvo.dec.com!e2big.mko.dec.com!pa.dec.com!decwrl!uunet!sun-barr!ames!sgi!rhyolite!vjs
From: vjs@rhyolite.wpd.sgi.com (Vernon Schryver)
Subject: Re: Point-to-Point Full-Duplex FDDI Link
Message-ID: <n0lflfs@rhyolite.wpd.sgi.com>
Organization: Silicon Graphics, Inc.  Mountain View, CA
References: <a563700.30.710527961@hp9000.csc.cuhk.hk>
Date: Tue, 7 Jul 1992 18:41:45 GMT
Lines: 23

In article <a563700.30.710527961@hp9000.csc.cuhk.hk>, a563700@hp9000.csc.cuhk.hk (Che-hoo Cheng) writes:
> Hi all,
> 
> I've heard that Digital supports Point-to-Point Full-Duplex FDDI mode on its 
> NIS brouter.  That means if you link up only two such nodes together 
> through point-to-point link, you can get 100Mbps throughput both ways.  My 
> questions are:  Is it a standard?  Do other vendors' products support such 
> mode?


Are you sure about that "100Mbps throughput"?  I hadn't heard the DEC was
close to 100Mb/s in one direction.
Note that I'm not talking about the raw speed of the photons, but the
number of user data bits moved from one machine to another.

Of course, there is no such thing as literally "Point-to-Point
Full-Duplex FDDI mode", because FDDI is a dual ring, not a star of FDX
links.  "Full-Duplex FDDI" is as impossible as "Full-Duplex ethernet."
It might be called something like "Point-to-Point Full-Duplex using
FDDI PMD and framing".


Vernon Schryver,   vjs@sgi.com
T.RTitleUserPersonal
Name		DateLines
637.1Full Duplex FDDI. Also, 100Mbits/sec. user dataMSBCS::REGEWed Jul 08 1992 15:4121


	Some internal performance studies have shown that in the TCP/Ip
	enviornment user data is transmitted at 89 Mbits/sec. on
	a VAX 9000 machine with DEMFA. I suppose that is close
	to 100mbits/sec.

	Users can never get 100Mbits/sec., because FDDI wire runs at
	100 Mbits/sec. and the user data has at least a minimal
	packet header. Infact the bandwidth available to a user
	is function of the packet size and all the headers with their
	sizes that are required to transmit the user packet.

	As you FDDI by definition cannot be full duplex. Full Duplex
	FDDI is a loosely coined term that allows a configuration
	of only two stations on a ring to use both the fibers, thus
	having the possibility of achieving 200 Mbits/sec.
	Note the packet format and other aspects of communication are
	same as FDDI and therefore the configuration is loosely termed
	Full Duplex FDDI.
637.2KONING::KONINGPaul Koning, A-13683Wed Jul 08 1992 16:3514
Apart from his snide remark about how slow we are, Vern's comment is technically
correct.  "Full duplex FDDI" is a term we've used to describe what would more
accurately be called "Full duplex point to point datalink using FDDI frame
formats".  The reason why it exists is that it gives you twice the available
bandwidth with very little change in the FDDI chips.  Of course, the bigger 
issue is what the increased bandwidth does to the adapter and the system.
I'm not sure that any of our adapters would handle the full duplex bandwidth.
The other question is how many of our protocol implementations can actually
consume more than 100 Mb/s.

Because of these two points, full duplex is mostly a placeholder for future
capabilities.  It may be useable today in some limited cases, though.

	paul
637.3JUMP4::JOYHappy at lastFri Jul 10 1992 07:127
    Much of the GIGAswitch public info and PID discuss use of FDX FDDI for
    communicating between the switch and a host, or between switches. I
    imagine that when GIGAswitch ships, we'll be seeing a more wide speread
    use of this capability.
    
    Debbie
637.4where is the racetrack ?MSBCS::KALKUNTERam Kalkunte 293-5139Fri Jul 10 1992 16:4412
    re .1
    The VAX 9000/DEMFA system can do 89Mbps for UDP, not TCP.
    
    I agree that FDX FDDI has no meaning. It is really FDX using optical
    cables, but we use the capability of already existent FDDI hardware.
    
    I believe sgi has seen throughput "close to 100Mbps" with UDP run
    without checksums, but Alex Conta's numbers for UDP include the
    checksums. Checksums cause a 20-30% hit in performance, so you
    know who is faster :-)
    
    Ram