Conference 7.286::fddi

    re .0
    
    
    With good planning/development the FDDI standard will ultimatly
    produce exceptionally fault tolerant networks, w/ transparent
    and auto reconfig capabilities. Not that theese "features" do not
    exist today. Additionally high speed/broader bandwith will reduce
    the bottleneck affect of some comm. links.
    
    A recent issue of Digital Review (Sept 25/89) had an article indicating
    the White House endorsed a National Supercomputer Fiber Optic Network
    It is a  $1.9 billion program to develop a fiber optic network to link
    supercomputers at universities ,and research centers across the
    country.  This Network would transmit data between supercomputers
    at 3GB per second. (Thats 2,000 times faster than existing networked
    computers). They also go on to say that this could be *finished* in the
    next 10 years.
                                  cheers/bob

    Well what "I" have heard at Net U and other seminars is that the
    "early" FDDI implementations will likely be "backbone" implementations. 
    That is: Ethernet will be the "workgroup network" and FDDI will connect 
    all the Ethernets together. Some "High Performance" Systems will 
    connect directly to the FDDI backbone. Fiber to the desktop (on a large 
    scale) won't happen for a while.

    I had expected FDDI to provide a fault tolerant network. However,
    in the recent information I've seen, I see Digital endorsing a Star
    wiring tology as this affords better tolerance in the event of multiple
    node failures. I don't consider this to be any more fault tolerant
    than todays networks where it is possible to loose a single building
    or user and leave the otehrs functioning.
    
    I'd like to see a method of wiring and hardware placement which
    could ensure full fault tolerance. I.E. A single cable or component
    failure could not cause any noticable degredation of network
    performance. This is highly critical to the distributed manufacturing
    and engineering environments.
    
    Since I work for Digital, I also expect Digital to present our
    customers with a method to utilize the DECconnect wiring they've
    installed for FDDI. Over the years we've been telling our customers
    that DECconnect would carry them into the next generation of LANs.
    I hope this includes 100mbit network access as in FDDI.

    Mike

    The kind of fault tolerance you are describing can only be achieved
    through full redundancy. That is, to prevent "any single component
    failure from making a node/building/sub-bnet unreachable", there
    musrt be a second net ready to take opver when the first fails
    at any point.
    
    There is no technology that could ever be invented to satisfy
    those requirements without full redundancy When something
    breaks,.. its broken, n'est-ce pas?
    
    				Bill Salkewicz
    				DECnet-VAX Development
    

Hi,
a couple of observations about what you said Mike.
Yes I agree about the fact that the Star Topology does not give full dual path
redundacny on all points of the network, however you do not see that many 
outages of Ethernet at the moment and by using multiple strands of fibre when
the cables are laid a break can be quickly patched out. As far as DECconnect
goes there is a document called DEFON (Digital Equipment Fibre Optic Network)
which gives information on configuring Fibre nets.
I suppose if you were really worried about redundancy down to individual
stations wire the whole thing as dual access stations. Very expensive and not
flexible.

Cheers Simon...

There is a certain amount of redundency built into FDDI.  Digital's wiring
concintrator DEFCN can be a "Dual Attachment Station" (DAS).  This means that
it connects to two other stations on the "Dual Ring" with a pair of fibers to
each. One fiber of each pair is a backup (or counter rotating ring) incase the
other pair of fibers, or station at the end of those fibers, fail.

In case of a failure, the station that can no longer transmit on the primary
ring will loop onto the secondary ring.  The data will be passed along the
secondary ring until a station that is no longer receiveing on the primary ring
takes the data on the secondary ring and loops it to the primary ring,
bypassing the failure.

The concintrator also has single attachment ports that go to a "Single
Attachment Stations" (SAS).  If there is a failure communicating on a port, the
DEFCN mearly bypasses the failing port.  Unless that port goes to SAS
concentrator, you loose only one station.

re DEFON
IT is DFON  "DECconnect Fiber Optic Network".

keith

Many FDDI vendors have been trying to make people believe that, just because
FDDI is a "dual ring", it is the whole answer to fault tolerance.  This just
is not even close to valid.

If you are serious about fault tolerance, you have to do failure analysis and
put redundancy (double, or maybe even triple) where needed.  It's just possible
that the FDDI dual ring will at times give you one of the pieces of redundancy
needed.  But usually you need more than that, and the dual ring may not
even help.

Putting significant dependence on the dual ring also introduces all sorts of
wiring and management problems.  Jerry Hutchison has an excellent paper 
(which he presented at the LCN conference early this month) on the subject.

FWIW, the US Navy used to think that dual rings were the answer.  They seem
to have changed their tune recently.  And of course they are among the more
serious users of fault tolerance... :-)

	paul

    The FT aspects of FDDI are certainly strong points, but I have other
    expectations...
    
    FDDI and OSI are effectively the first protocols that began as
    standards, as opposed to proprietary vendor implementations that were
    later standardized (e.g. Ethernet => 802.3, Token Ring => 802.5).  So
    no vendor will have the claim that "we invented it", and this will
    automatically make it more appealing to a wide range of customers that
    are very tired of proprietary implementations that will not
    inter-connect.
    
    Along these lines, I expect that FDDI will become a multi-vendor,
    multi-protocol backbone that will assist in unifying the disparate
    topologies of today, particularly 802.3 and 802.5.
    
    In addition, of course, the high speed and low-noise fiber environment
    will lead to the implementation of more networked applications along
    the lines of client-server, X-windows, database servers, etc.  I think
    DEC is very well positioned to be a major supplier of this technology,
    and will be viewed as a standards-based, progressive technology company
    for the first time to many of our customers.
    
    Lest you think I have only grandiose and noble expectations, I also
    think that FDDI will sell like hotcakes in late FY90 and FY91, and make
    most of my budget...