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

Re: Data rate standards vs internal switching standards

Roy Bynum wrote:

Yes, I am avoiding a question that is too early to answer.  The single biggest issue to me is that the majority of the group does not understand what extended and leased fiber systems and Native data communications services are, and what is required to do reliable implementations using them.

To date, I know of no HSSG objectives to address leased fiber systems.

I don't understand what you mean by "Native data communications services". Please explain. How is this relevant to the HSSG?

In order for the 802.3 committee to understand these issues, they must be educated. Those that wish to educate 802.3 should step forward and ready tutorials and presentations before 10 Gbps Ethernet objectives are cast.

 At the end of your tutorial on MAS, I attempted to explain a major potential use and need for MAS.  It was to determine the specific power requirements for spans of leased fiber.  Your statements below indicate that you did not understand what I was trying to say.  This in not unusual.  I am very glad that you are learning more about SONET/SDH and what commercial service providers do to provide reliable services.  Perhaps a broader understanding of the environment that 10GbE will find a major market in, will be achieved at the September meeting.
My statement in response to your comment is immediately below. MAS technology trades off SNR with symbol rate. However, the Transcendata MAS proposal uses the same optical power for a single symbol, that symbol carrying 2 bits of information. I am assuming that the specific power requirements will be the same on leased, rented or owned fiber.

A couple of questions: What is the purpose and relevance of "detecting inconsistent attenuation at splice points in leased fiber facilities in GbE", why should 802.3 care, what should we do about it and what should MAS do differently than binary signaling systems?

>> The presentation at the last meeting of the Multi-level Analog Signaling technology recognizes
>> that there was no provision for detecting inconsistent attenuation at splice points in leased fiber
>> facilities in GbE.

> Please elaborate on this point also. I don't recall a MAS slide related to this issue.

You acknowledge that Ethernet and fibre channel technologies do not have the ability to support any cable or fiber facility that is not directly owned by the user, that includes 1000BaseLX.
Ethernet and Fibre Channel technologies support cable and fiber facilities according to their respective standard specifications. These specifications make no distinction between leased, rented, owned or even stolen fiber. I view a 1000BASE-LX link which utilizes leased fiber as a 1000BASE-LX link. I'm not sure what you're getting at, but if you're insinuating that a 1000BASE-LX link over leased fiber MUST include SONET, SDH, ATM, etc., then I STRONGLY disagree with your statement.
 What you may not be aware of is the amount of leased fiber implementations of 1000BaseLX (GbE) and the push in the market for providing Native data transport services.
That sounds great for the Gigabit Ethernet market!
 It is my understanding that this forum does not allow referencing company's names otherwise I would refer you to several GbE that are implementing such systems for their customers.  I will refer you to the July 20 Network World article about a school system that has implemented a MAN using fiber leased from a utility company.  This is just one of those enterprise customers that are looking at optical networking as an alternative to traditional data telecommunications services.  The amount of leased fiber and wavelengths that will be available to enterprise customers over the next few years is staggering compared to what was available last year.  This is a major potential market for 10GbE.
This is all great news. I agree with all of it!
The combinations of long haul Native data transport, leased fiber, and leased wavelengths will provide major expansion of the role of 802.3 in the market place.   The role of switching 802.3 instead of routing in the enterprise network space will be expanding dramatically in the next few years.  If it is structured properly, 10GbE will have a major role in that.  If it is not structured properly, it will not.
I for one intend to insure, to the best of my ability, that 10 GbE is structured properly to ensure the satisfaction of existing markets and the expansion into new markets.
 Until more people that are in HSSG, understand what the implementation requirements are for the customers using the expanded services and their service providers, I will attempt to avoid the issue of a compromise on a variable IPG.
I and many other in 802.3 are waiting with baited breath to hear about these implementation requirements.
Thank you,
Roy Bynum
MCI WorldCom
Best Regards,


Rich Taborek wrote:

It's pretty clear that you're avoiding the question now posed to you over this reflector about a half dozen times. Here it is again:

>Recently, Ariel Hendel of Sun has proposed a 10 Gbps rate solution along with a
> variable IPG. I consider this to be a reasonable compromise (not my best choice)
> to settle this outstanding objective. I have still not heard any good technical
> reasons as to why this solution would not work. Would you care to offer clear,
> succinct reasons to this effect?

I'll indulge your tangential thread anyway.

Best Regards,


Roy Bynum wrote:


The fibre channel standard is predispositioned toward self-owned fiber facilities.

So is Ethernet, currently. In the near future (3-5 years) a very small percentage of Ethernet interfaces may permeate the leased fiber space.
ESCON reliability is very dependent on being able to control access to, and
potential ad hoc damage to, the fiber facilities by unknown third parties.
Yes. As is Ethernet and Fibre channel because the fiber facilities are "self-owned".
There are no provisions for cross fiber systems transport in fiber channel.
I don't know what this means. Please explain.
The presentation at the last meeting of the Multi-level Analog Signaling technology recognizes that there was no provision for detecting inconsistent attenuation at splice points in leased fiber facilities in GbE.
Please elaborate on this point also. I don't recall a MAS slide related to this issue.
 There are no provisions for ad hoc damage to fiber facilities by unknown third parties. OIF has recognized that in order to transport fibre channel protocols over standard optical networking systems requires an out of band "digital wrapper".  This is not my "take" on this issue. This is the reflected in the considered opinion of the entire optical networking services industry.
I coming to understand what SONET is and am leaning every day. I believe you need to understand that Ethernet solves most of the same problems that SONET does in a different, simpler, and more cost effective fashion.
Thank you,
Roy Bynum
MCI WorldCom

Rich Taborek wrote:

> Roy Bynum wrote:
> > Rohit, et al,
> >
> > What is to prevent some unscrupulous vendor from sell a non WAN/MAN support
> > capable version of 10GbE to someone?  Are we to be responsible for adding to
> > the "caveat emptier" atmosphere of this industry?  As it is, a lot of the
> > sales and marketing people in this industry have the habit of advertising
> > and sell functionality that they do not have.  Are we to become part of
> > that, by creating a "cheap" version of a protocol that causes support
> > problems for customers?
> Roy,
> I'm not aware that Gigabit Ethernet, Fibre Channel, IBM ESCON(TM), etc. are
> perceived as "cheap" protocols. On the contrary, my perception is that they are
> very cost-effective protocols, much more so than their WAN counterparts. I have
> pointed out in a previous note that LAN and SAN links of significant distance
> (up to 10's of km) are widely deployed in LAN and SAN applications.
> Relative to customer support problems: the Gigabit Ethernet Physical Layer is
> heavily based upon the Fibre Channel Physical Layer, which in turn is heavily
> based upon the ESCON Physical and Signaling layers. I have personally worked on
> all three. Most of the same facilities to provide link fault detection as well
> as protocols capable of providing fault tolerance and isolation exist in all
> three. Most mission critical transactions in the USA and a very large portion
> worldwide still employ ESCON channels for reliable data transport with 100% data
> integrity and zero downtime. Whatever part of a transaction goes out on a WAN is
> still guaranteed by protocols at a higher level than the optical transport. I
> can make a pretty good case for GbE and its follow-on, 10 GbE being as reliable
> as ESCON and far more cost effective (primarily due to volume).
> Just because the LAN and SAN camps approach customer support from a different
> viewpoint than does the WAN camp, please don't jump to the conclusion that their
> solution is inferior. It is CLEARLY orders of magnitude more cost effective. I
> know that I don't fully understand the WAN solutions for reliable data
> transport, but I am interested in a serious discussion of the inadequacies of
> Ethernet in meeting customer support objectives.
> > The whole reason for the proposal of a 9.584 MAC transfer rate was to allow
> > the standardization of an optical 10GbE protocol that has the support
> > functionality that is recognized by the optical networking industry.  There
> > are several people on this reflector that are also participants in the
> > Optical Interoperability Forum (OIF) that was started by data communications
> > vendors, among others.
> Many very experienced and highly respected 802.3 members have proposed several
> solutions to 10 vs. 9.58464 MAC/PLS rate objective.
> Recently, Ariel Hendel of Sun has proposed a 10 Gbps rate solution along with a
> variable IPG. I consider this to be a reasonable compromise (not my best choice)
> to settle this outstanding objective. I have still not heard any good technical
> reasons as to why this solution would not work. Would you care to offer clear,
> succinct reasons to this effect?
> P.S. My choice is to use 802.3x flow control between the two rates and configure
> the network to prevent/minimize packet loss.
> > I joined this group for the specific purpose of preventing the spread of an
> > optical protocol that belonged on copper, not on fiber.  I have no issue
> > with 10.0 MAC transfer rate for copper, it can not be implemented of 100s
> > and 1,000s of km where the user/customer does not have viability or ability
> > to support he fiber that he is using.  Any protocol that can be implemented
> > over optical fiber can and will be used over fiber that is leased, not
> > owned.  That means it can not be supported without built in OAM features in
> > the protocol.  Those features are a REQUIREMENT as far as I am concerned.
> My interpretation of the statement above is that you disagree with all current
> HSSG objectives since those objective do not currently support any copper
> variant. Is this true?
> Ethernet has supported optical PHYs of significant distance at data rates of 10
> Mbps, 100 Mbps and 10 Gbps. It is only natural that optical PHYs of similar
> distances are specified for 10 Gigabit Ethernet. In addition, based on
> Ethernet's expanding markets, it is appropriate to extend optical PHY distances
> for 10 GbE. This is clearly reflected in the 10 km and 40 km singlemode fiber
> optical PHY HSSG distance objectives.
> SONET OAM features have never even been proposed as a possible candidate as an
> HSSG objective. I would suggest that if you're interested in this functionality,
> that you make a formal presentation at the September HSSG meeting. My best
> judgement is that the functionality of SONET OAM features can more cost
> effectively be met by standard Ethernet features such as Management and
> Auto-Negotiation.
> > The use of 9.584 transfer rate has been proposed as the most cost effective
> > and simplest way to provide a base for adding OAM features to the optical
> > PHYs of 10GbE.  It leverages the existing SONET/SDH technology.  It may be
> > that some vendors would like to be able to have their intellectual property
> > be part of the 10GbE standard.  As a data network architect, designer,
> > implementor, and customer, I would rather see a public domain standard.
> The simplest, most cost effective and open means of providing 10 Gbps Ethernet
> transport is by leveraging all that Ethernet is and doing it ten times faster.
> > Thank you,
> > Roy Bynum
> > MCI WorldCom
> >
> > Rohit Sharma wrote:
> >
> > > I concur entirely.
> > >
> > > "KISS" and make up...
> > >
> > > -rohit
> > > Rohit Sharma
> > >
> > >
> > > > -----Original Message-----
> > > > From: Mike Bennett [mailto:mjbennett@xxxxxxx]
> > > > Sent: Thursday, August 05, 1999 12:46 PM
> > > > To: IEEE HSSG
> > > > Subject: Re: Data rate standards vs internal switching standards
> > > >
> > > >
> > > >
> > > > I think its time the WAN folks consider a call for interest
> > > > to do the SONET
> > > > compatible standard separately from the LAN/MAN efforts.
> > > > This would allow
> > > > consensus for a 10 Gbs MAC and we could move on.  Devendra
> > > > Tripathi alluded
> > > > to this yesterday, although to give it an "honorable mention"
> > > > rather than
> > > > make it standard.  I'm curious if others would support a
> > > > separate effort in
> > > > hope that both groups could move forward.  I'm not opposed to a WAN
> > > > standard, but it seems as though there is no end in sight for
> > > > a compromise,
> > > > that is to choose either a 10 or Gbs data rate.  Some seem to be
> > > > willing to sacrifice the simplicity and low cost of Ethernet
> > > > to get into
> > > > the WAN market.  In my humble opinion as an end user of
> > > > Ethernet equipment,
> > > > if we stray from that which has made Ethernet so successful to date it
> > > > would negatively affect the future of Ethernet.
> > > >
> > > > Mike Bennett
> > > >

Richard Taborek Sr.    Tel: 650 210 8800 x101 or 408 370 9233
Principal Architect         Fax: 650 940 1898 or 408 374 3645
Transcendata, Inc.           Email: rtaborek@xxxxxxxxxxxxxxxx
1029 Corporation Way    
Palo Alto, CA 94303-4305    Alt email: rtaborek@xxxxxxxxxxxxx