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

RE: Distance Motion


802.3 is already being encapsulated into SONET systems by SONET
systems vendors. This is a currently a limited market space due more
to a paradigm issue than technology. At present, GbE over OC192 is
available and will start being deployed sometime next year. With GbE
over OC192, 10XGbE is achievable today. GbE tributary interfaces in
the OC192 SONET transport LTE are less expensive per bandwidth than
comparable SONET tributary interfaces. It is still more expensive than
a 10GbE (OC192C) interface, that directly interfaces to SONET systems
from data switches will be. Directly incorporating the 10GbE (OC192C)
interface as part of the 802.3 standard will greatly reduce the
overall cost of this implementation. This will allow implementation of
high bandwidth WAN systems within the existing transmission facilities
With little or no extra cost to the carriers, and a reduced cost to
the customers. 

I don't know where the idea that more re-tooling will be required to
support a 10GbE (OC192C) rate than any other version of 10GbE. Other
than the inclusion of additional L1 WAN PHY functionality, there is
not different between this and any other implementation at that rate.

It is the WAN operations and maintenance functionality that is needed.
Until it was suggested to me by a vendor, I was not even thinking in
terms of making it an OC192C rate. The suggestion solves a whole lot
of issues with having to build a unique transmission network for
802.3, instead of leveraging the existing technology in the existing
transmission network.  If that can save money in a serial non-WAN
implementation, the so much the better.

With all of the concern about BER, it would be good to have BER indicators
in the 10GbE PHY.  This is just one of the many functions that the
SONET/SDH overhead provides transmission systems.

Roy Bynum
Optical and Data Network Technology Development
MCI WorldCom

Date:     Fri Jun 25, 1999  3:38 pm  CST
Source-Date: Fri, 25 Jun 1999 17:38:47 -0400
From:     "Chang, Edward S"
          EMS: INTERNET / MCI ID: 376-5414
          MBX: Edward.Chang@xxxxxxxxxx
TO:     * ROY BYNUM / MCI ID: 424-5935
TO:       "Chang, Edward S"
          EMS: INTERNET / MCI ID: 376-5414
          MBX: Edward.Chang@xxxxxxxxxx
CC:       HSSG Distance
          EMS: INTERNET / MCI ID: 376-5414
          MBX: stds-802-3-hssg-distance@xxxxxxxx
Subject:  RE: Distance Motion
Message-Id: 99062521382826/INTERNETGWDN1IG
U-Content-return: allowed
U-X-Mailer: Internet Mail Service (5.5.2448.0)

Nice to have a interesting discussion with you again. 

You are very knowledgeable in all types of network, LAN/MAN/WAN, and we are
fortunate to have you in our team to help us.

I myself is mainly LAN type person; however, I also worked on several WAN
related Unisys projects to understand WAN enough to appreciate your
concerns. I think the reason LAN people resist whole blown embrace of SONET
protocol into 10GbE is not because your idea is wrong, rather than it is
very expensive to wholly embrace SONET protocol and OAMP into 10GbE.  The
engineer, field service, chip makers, software, all have to be retooled in
one way or another.  It will take several years for LAN industry to absorb
that.  Furthermore, there are new issues:
the compatibilities are imposed into each LAN terminals, instead of being
dealt at SWITCH/ROUTER/BRIDGE level.  All of these will make all Ethernet
family products too expensive to compete in the LAN market, if we adopt
Sonet OAMP...etc.

I would ask how expensive it is to encapsulate, or map 10GbE into SONET.
FDDI has been doing encapsulation to SONET for a long time, they like it.
The popular TCP/IP can capsulate many different protocols with amazing
popularity; therefore, with so many different protocols in the industry,
encapsulation is one of the most money making business now -- TCP/IP
software engineers are hot.  Probably we should propose a task force to
study how to encapsulate 10GbE into SONET effectively to ease the pains of


Ed Chang
Unisys Corporation





-----Original Message-----
From: Roy Bynum [mailto:RBYNUM/0004245935@xxxxxxxxxxx]
Sent: Friday, June 25, 1999 2:01 PM
To: Chang, Edward S
Cc: HSSG Distance
Subject: RE: Distance Motion


I have been building and supporting different types of data network
architectures for many years. I built my first LAN in 1982. I built my
first WAN in 1986. I build my first MAN in 1991. I have been active at
one level or another of all of the data networking architectures and
implementations within MCI for all of the 10 years that I have been
with this company. I am the first registered host master for MCI. I
designed the architecture for the original common access routed WAN
for MCI. Variants of that architecture are still in use today. I
designed the architecture for the internal data communications
networks, surveillance, Intelligent network services, back office, and
administrative services, for the Avantel telephone company in Mexico,
including the LANs, MANs, and WANs. I am currently part of the
technology evaluation and definition group for MCI Worldcom, including
future data networking technology, internal and external offerings.

A major portion of the economic impact of 802.3 is the lack of
operations complexity; it does not have/require very much in the way
of configuration/operations maintenance for it to be operable. This
works very well within well defined and easily accessible facilities.
It has gotten massively better as the technology has moved from the
original cable systems to non-blocking switched twisted pair systems.
As the facilities become extended over multiple sites, the lack of
operations maintenance functionality starts to become an issue.
Additional support people for each site are needed to have timely
response to support issues. The alternative to additional support
people is excessive reserved bandwidth, or accepting long downtime
periods when sporadic equipment and fiber issues occur. The support
portion of operating costs becomes even greater as the "LAN" extends
out to what used to be "WAN" environments.

The reason that there is such a confusion as to the distinction between
a MAN and a WAN is that it has, up until now, been in the purview of
the telephony carrier industry.  The difference between a MAN and a
WAN is that a WAN spans central office facilities, while a MAN originates
and terminates in the same central office.  Most of the time, this
can be co-related to distance.  A WAN is generally over 30-50km. A
MAN generally is less than 30-50km, or the length of two local loop
services from the same central office.

Few people, if any, see the operations maintenance functionality that
exists in the traditional MAN/WAN, T1/DS1, T3/DS3, or SONET rate
systems. Even the lowly T1/DS1 has one bit in every 194 reserved for
operations support. The T3/DS3 has 1.7Mbps reserved for various
operations support functions. This built in, fixed, overhead allows
MAN/WAN architecture designers, to implement data networks without
having to directly incorporate the "operations" support, for the L1/L0
functionality, in the "data" layers of the network. The built in
operations support allows the traditional MAN/WAN to have circuits
that work, without having to figure out where in the transmission
system itself any issues might be originating. Currently all of the
transmission system issues are addressed by the carriers which use the
built in operations functionality to reduce the cost of the services to
the end users. From a basic functionality viewpoint, the presence or
lack of built in operations support is the basic difference between
MAN/WAN L1 protocols and LAN L1 protocols. 

Because of a lack of operations support functionality, 802.3 requires
a higher level of talent and expertise to support over MAN/WAN
environments than does traditional MAN/WAN systems. That talent pool
is limited and expensive. In time, it will be as expensive to support
an 802.3 MAN/WAN, as it does to support an ATM MAN/WAN, if not more
so. ATM, at least, has a SONET transmission adaptation layer. The
current 1000BaseLX standard does not even have that. Can you imagine
what it will be like as more and more companies/people deploy 802.3
GbE MANs and WANs. Right now, GbE is in the "honeymoon" of a new,
capital inexpensive technology. What will it be like when the
"honeymoon" is over? It is going to turn into a support cost nightmare.
Because of that, GbE will not have a major, long term, deployment in
direct optical MAN/WAN systems. Instead, GbE will be encapsulated,
"mapped", into SONET transmission systems. While the cost of the SONET
"mapped" costs will be higher than the direct optical transport costs,
it will at least be serviceable, reducing the overall cost of ownership
of GbE MAN and WAN systems.

It is the long term implementation issues that I am concerned about
with 10GbE. We have the opportunity to develop a standard that provides
built in operations functionality. That functionality will allow 802.3
MAN/WAN implementers to provide systems that are as cost effective as
other 802.3 LAN systems are. It will allow 10GbE to have a broader,
longer term, deployment in MAN/WAN environments than GbE will have. We
have the opportunity to leverage existing optical transmission
technology that already contains proven operations support
functionality. Re-thinking the role of 10GbE in the MAN/WAN
environment will have a major impact on its success long term.

Thank you,
Roy Bynum,
Optical and Data Network Technology Development,
MCI WorldCom

Date:     Fri Jun 25, 1999  9:52 am  CST
Source-Date: Fri, 25 Jun 1999 11:49:59 -0400
From:     "Chang, Edward S"
          EMS: INTERNET / MCI ID: 376-5414
          MBX: Edward.Chang@xxxxxxxxxx
TO:     * ROY BYNUM / MCI ID: 424-5935
TO:       Thomas Dineen
          EMS: INTERNET / MCI ID: 376-5414
          MBX: tdineen@xxxxxxxxxxxxx
TO:       HSSG Distance
          EMS: INTERNET / MCI ID: 376-5414
          MBX: stds-802-3-hssg-distance@xxxxxxxx
Subject:  RE: Distance Motion
Message-Id: 99062515523740/INTERNETGWDN3IG
U-Content-return: allowed
U-X-Mailer: Internet Mail Service (5.5.2448.0)
HSSG Distance Friends:

Such a interesting discussion, I just could not resist to jump in to join
the crowd.  

I believe our objective is to determine the distance specification for 10GbE
which basically is LAN.  As in many LAN standard generations (Ethernet,
FDDI, ATM, Token Ring,..), and even Fibre Channel, the distance issue has
been determined based on the Cost-Performance, the Market Needs, but not
types of "Components".  I just wonder why type-of-component became the main
focus for our distance selections.  Furthermore, why we should be
responsible for LAN, MAN and WAN definitions, considering our task is to
provide 10GbE users the capability of operating distances using any
technology of the most cost-effectiveness.  

Just to make sure I have the correct understanding of technical terms, I
reviewed several Telecom, Datacom, Communication...etc. Standard Reference
publications in library about the definition of LAN, MAN, and WAN.  None of
them define LAN, MAN, and WAN based on the component types.  The only clear
statement common to different reference books is "WAN typically connects
many LAN clusters, and WAN is subject to public telecommunications
regulation", but not LAN.  MAN definition is so vague, which can have both
of them as long as it is in the middle. 

I am involved, and make technical decisions for Unisys world wide optical
communications, which are LANs basically.  We have inquiries from users all
over the world to extend distance beyond their present installations ranging
100 meter to over 50 km.  The only thing market cares is performance and
cost, but not technologies or types of components.  By the way, SONET uses
LED and MM fibers for interoffice connections for very low data rate. 

From LAN/WAN-Standards and Market-needs point of view, there are no
regulations that constraint us from making DISTANCE decisions.  I believe
the only rule we have is our obligation to users, to marketplace to provide
the most cost-effective distances for 10GbE. 

Ed Chang
Unisys Corporation


-----Original Message-----
From: Roy Bynum [mailto:RBYNUM/0004245935@xxxxxxxxxxx]
Sent: Thursday, June 24, 1999 6:09 PM
To: Thomas Dineen; HSSG Distance
Subject: Distance Motion

Thomas, et al,

I am in favor of having a LAN only distance definition, motion item

Because of optical networking issues, such as optical amplifiers, OEO
transducers, DWDM, etc., I can not seperate the MAN (10000BaseLX)
interface from a WAN (10000BaseLX) interface. In spite of an attempt
to create a language that appears to limit the distance to MAN
applications, exiting 1000BaseLX is being deployed as a WAN interface.
In optical transmission interfaces, there is not a distinction between
MAN and WAN. They use the same wavelength; they use the same output
power; they use the same framing and link protocol; they use the same
single mode fiber; they use the same physical interface. Look at what
Bill St. Arnaud is doing at CANARIE. I can not agree to the motion to
seperate the MAN from the WAN, items 2 and 3.

The issue of fiber type and distance are specifically linked. Any
interface that uses multi-mode fiber (MMF) is limited to LAN systems.
Any interface that uses single-mode fiber (SMF) can be used in MAN/WAN
systems in addition to LAN systems. In MAN/WAN systems, wavelengths
can be either in the S band, C band, or the L band. All three use SMF.
All three have been used for interfaces in both MAN and WAN systems.
The difference that I see between LAN and MAN/WAN interfaces is the
fiber type.

Roy Bynum
Optical and Data Network Technology Development,
MCI WorldCom
(972) 729-7249