Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Distance Motion




Jonathan, et al,

I laud your motion to alter the objective definition of "distance" for
10GbE.  This has become a major hot topic for fiber optic systems
implementors, both in metro and long haul services.  The advent of
active and intelligent DWDM systems has compounded the confusion.

The issue of distance is one of understanding the difference between
what is a LAN and and what is not a LAN.  While I may be somewhat
archane, I have always understood that a LAN was restricted to a
building.  When it left the building, it was no longer a LAN, it was
something else.  It did not mater if you were bridging or routing, if
the data system left the building, the part outside the building was not
part of a LAN.  Just because a broadcast domain could be extended over
great distances by means of bridges did not make that longer distance a
LAN.  What existed over the non-LAN was a virtual segment, not a real
one.

Because of a lack of understanding the implications of optical
networking, 1GbE included a long haul wavelength and power
implementation, as an extension of the LAN standard.  It exists in the
telephony standards of Intermediate Reach.  Optical networking provides
for direct conversion of telephony intermediate reach to telephony
standard long reach simply by using an optical transducer.  In spite of
what was intended or how it is spoken of; 1000BaseLX is not a LAN
standard.

In order for 10000BaseXX to be a LAN standard, it must be restricted to
a 850nm, single mode fiber, PHY.  This would allow the 12.5Gb signal
rate for 8B/10B block code within the confines of a building.  This
standard would be a "10000BaseSX".

On the other side of an L2 data switch, for non-LAN systems, long-haul
optical networking issues must be considered.  This is an environment
that has different operational support requirements than a LAN.  Getting
access to the fiber plant is not as simple as tracing it through the
ceilings, floors, or risers of a building.  In some cases, you will have
to dig up the street, to get access to non-LAN optical fiber.  That
makes the support costs of non-LAN optical transmission systems a
totally different economic and business model.

The economic and business model of GbE in the MAN and WAN environment is
based on the lack of conversion from one L2 protocol to another.  L2
switching has a major econonomic saving compared to L3 switching.  The
common equipment and interface costs for L2/L3 switches is .10 to .25
the costs of L3 only switches.  A major portion of the additional cost
is the processing required to strip the L2 protocol from the L3 and then
put it back on again.  Because of the lack of L2 conversion, today's
L2/L3 switches have a much higher density and economic savings than any
L3 only switch, and compare very favorably with the new generation of
"terabit routers" being developed.  The only problem is that there is no
support for operations of the optical fiber plant in the GbE protocol.

A seperate "long-haul" version of 10GbE will allow continuence of the
L2/L3 switch interface and common equipment economic model, while
providing operations support for non-LAN implementations.  This would
make a distiction between the short-haul, MMF, 850nm standard and the
SMF, intermediate/long-haul, 1300/1500/1600nm standard.  This will allow
for different signalling standards; block encoding for LANs and scramble
encoding for MANs/WANs. It will also leverage the existing SONET/SDH
installations and experience of the common carriers. 

Personally, I think that the addional 25Gbps signalling rate will cost
more than the savings gained by block encoding.  When the issue of
launch power is removed, I think that a SONET/SDH signaling PHY will be
less expensive for LAN implementation than an 8B/10B signaling PHY. 
Having seperate standards will also allow for economics, not personal
viewpoints to define the prevalent implementation for the different
envionments.  

Thank you,
Roy Bynum
MCI WorldCom