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

RE: more on : IEEE 802.3 Requirements




Roy:

Thanks for comments.

I appreciate your result of economic study, which indicates GbE is extremely
cost-effective, than other technologies.   

Cost-effectiveness is always the top priority for all Ethernet family of
products including 0.01G/0.1G/1G/10G rates.

The 10 GbE is no exception.  I believe some vendors are aiming for the 10GbE
per port price of about three times of the GbE one port price.  The Ethernet
market will force the 10GbE 
price to be in line with the traditional Ethernet price setting -- very
cost-effective.  I firmly believe the short haul 10GbE is no exception,
which will be very cost-effective.  

Ed Chang
Unisys Corporation
Edward.Chang@Unisys.com  



-----Original Message-----
From: Roy Bynum [mailto:rabynum@airmail.net]
Sent: Tuesday, May 25, 1999 8:37 AM
To: Chang, Edward S
Cc: elg@cypress.com; stds-802-3-hssg@majordomo.ieee.org;
msalzman@lucent.com
Subject: Re: more on : IEEE 802.3 Requirements



Edward,

In doing economic studies of GbE versus other systems, it was found that
10 GbE ports with DWDM was cheaper than equivilent ATM or OC48 POS. 
Even with the higher number of ports and the additional cost of DWDM,
the overall port and common equipment costs were less for GbE.  For
someone to implement 10GbE, it will have to be more economical than
GbE.  Has anybody considered the economics of these various proposals
for short haul 10GbE?

					Thank you,
					Roy Bynum


Chang, Edward S wrote:
> 
> Some preliminary data to support 10 GbE over MM fibers.
> 
> By using the "restricted launch"(optimize shortwave link), "off-set mode
> control"(optimize longwave link), and "retry protocol" (weed out bad
> fibers), we can take advantage of using the installed MM fibers for 10GbE
> applications.  The bad fibers can be either left unused, or replaced by
new
> MM super-fibers to be introduced by vendors.  At 10 Gbps, we may not need
as
> many links as 1 Gbe, and 100 Mbps anyway -- use less links.
> 
> In order to have the feeling of if using the installed base is worthwhile,
> we can estimate the
> 10GbE performance (62.5 um MM at 10^-12 BER) from the installed GbE data:
> 
> 1. Longwave (1300 nm) links at 10 Gbps.
> 
>         At 100 meter link length -- almost all fibers will work, provided
a
> "mode-control" is used  as it is needed.  The rejection ration is very
> small, perhaps, only several %.
>         At 250 meter -- probably, over 60% of the installed FDDI grade
> fibers will perform.
> 
> 2. Shortwave (850 nm) links at 2.5 Gbps for WDM applications.
> 
>         At 200 meter -- Most of fibers will perform, provided   the
optimum
> "restricted launch"     transceivers are used.  The rejection ratio should
> be only several %.
>         At 400 meter -- perhaps, over 70% of fibers will perform.
>         At 1 km -- perhaps, over 40% will perform
> 
> 3. Shortwave (850 nm)links at 10 Gbps --- laboratory data showed VCSEL can
> perform at 10 Gbps.
> 
>         At 50 meter -- most of the installed fiber will perform, only
> several % of rejection.
>         At 100 meter -- perhaps, over 70% will perform.
>         At 250 meter -- perhaps, over 40% will perform.
> 
> Depending on how much we like to remove those defected fibers, we can set
> different minimum length specifications.  All those to be removed, bad
> fibers are DMD fibers, which have different degrees of DMDs from severe,
> obvious to minor, not-obvious.  They affect the bandwidth reduction
> differently, causing very low-BW fibers to fair BW fibers.
> 
> We can use the existing "retry protocol" to identify those bad, DMD,
fibers
> without any added product cost.  Furthermore, we can train servicemen to
> remove those bad fibers at the first installations.
> 
> As long as we update the fibers systematically, and orderly, the updating
> will provide better cost-performance products for users, and stimulate
> business for vendors.  Look! Microsoft, Intel, PC, Memory, Internet ......
> keep updating.  The results -- users are happy, more efficient, and
vendors
> keep raking-in more business.
> 
> Among cables, the copper cables have been updated so many times to meet
the
> needs of the more efficient technologies; however, the MM fibers are still
> using the fibers developed in the FDDI era -- more than 15 years ago?  I
> believe, it is time to update those old fibers to provide  better
> cost-effective media for users, and opportunities for vendors to create
more
> business.
> 
> Ed Chang
> Unisys Corporation
> Edward.Chang@Unisys.com
> 
> -----Original Message-----
> From: elg@cypress.com [mailto:elg@cypress.com]
> Sent: Thursday, May 20, 1999 10:44 AM
> To: stds-802-3-hssg@majordomo.ieee.org; msalzman@lucent.com
> Subject: Re: more on : IEEE 802.3 Requirements
> 
> Hi Mike,
> 
> While you may be uncomfortable with it, that response is similar to
> what 802.3ab implemented.  With respect to the installed base, just
> how much of it must work correctly?  99.999%? 99.9%? 95%?  Just where is
> the cutoff point that makes the interface not good enough for public
> use?
> 
> As far as testing, right now the UTP users test all their links to see
> if they meet the link requirements.  That can be done because
> A) the test equipment is cheap
> B) if you don't test, your link may not work
> 
> With respect to operation on the installed base, there is a litany
> of the progress of 802.3 where the installed base WAS abandoned.
> If you don't believe me, show me how many 100TX installations are
> working on CAT3.
> 
> Even the magic being wrought by 802.3ab will not carry everything forward.
> There are numerous installations of CAT5 that will not work for 802.3ab.
> Some of these cables were only made with two pairs with a good dielectric,
> and two pairs with PVC (to save cost).  These cables met all CAT5
> requirements when they were installed.
> 
> Even with good cables, you still have other installation issues; i.e.,
> were the cables stressed during install, were CAT5 connectors used,
> were the pair twists maintained right up to the connector, etc.
> 
> You get the same with fiber.  When they pull a bundle, not all fibers
> are good.  Some may not pass any light at all.  I think there is nothing
> wrong with using the installed base IF (a big if) we can test the
> fiber to certify its capabilities.  If you want to use a specific fiber
> for laser use, you need to test it with laser launch, and not rely on
> a 20 year old test spec based on overfill LED launch.
> 
> Many of these fibers are just fine for the application at hand.
> Many are very bad.  I don't think the usage is out of the question
> for the installed base.  When it really needs is an economical
> and simple to use field test to allow the fibers to be characterized.
> 
> -Ed Grivna
> Cypress Semiconductor
> 
> >
> > Hi Ed,
> >
> > I am uncomfortable as a standards body telling people that they should
use
> > fiber X to reach 300M and to then say, well if you want to run over
> > installed base, try it and see how far you get.  We know that serial
> streams
> > at 10G or thereabouts, on 160MhzKm fiber will not even reach 100M.
> > Realistically, some fibers out there will go to 150 or more.  But,
> consider
> > that the buyer will go to some trouble to set up a new, expensive
trunking
> > facility and they will schedule some down time to deal with it,
> interrupting
> > many users in the process.  It seems to me unlikely that a project
manager
> > in that position will want to rely on unknown properties of of current
> > cable.  Furthermore, as the plant continues to age, what parameter will
> > fail, how and when?  We already know that existing fiber is simply far
> from
> > capable of supporting the new rates.  It is not a matter of close but no
> > cigar.  It is more than an order of magnitude difference in quality of
the
> > fiber.
> >
> > Therefore, my sense is that the project owner will decide up front
whether
> > to replace the fiber or not.  My guess is that on main trunks, they will
> > replace them with single mode.  In many new applications they could put
in
> > high performance multi mode.  In impulsive applications (which will be
> > scarce in the beginning) they will simply try to see if they can live
with
> > the installed base.  If a MAS or CWDM PMD is available, they might
simply
> > decide to try that one for their installed base applications.
> >
> > To your point about educating the masses of installers, experience
teaches
> > that it takes on the order of 5-10 years to inculcate an appreciable
> > quantity of them about the stringent aspects of installations.  How many
> of
> > them really know the differnces between cables and various kinds of
> fibers?
> > how many still do not know the total panoply of procedures and equipment
> to
> > install Cat5 correctly?  I think we want our technology in the market
> before
> > much of this knowledge will become commonplace, therefore we need to
> > enunciate a clear set of requirements in a black and white manner for
> these
> > people to follow.
> >
> > mike
> >
> > > -----Original Message-----
> > > From: Chang, Edward S [mailto:Edward.Chang@unisys.com]
> > > Sent: Wednesday, May 19, 1999 06:29
> > > To: msalzman@lucent.com; stds-802-3-hssg@majordomo.ieee.org
> > > Subject: RE: IEEE 802.3 Requirements
> > >
> > >
> > > Mike:
> > >
> > > You are right, I agree with your reasoning.
> > >
> > > The key points we are addressing is we should be flexible, convenient
to
> > > users, easily adaptable by market.  The rest of the headaches are upon
> > > technical people(us) to make it happen.
> > >
> > > We will allow the installed bases to be used without modifications at
a
> > > shorter distances -- perhaps, 100, 200, or even 300 meters?, then
> > > we can add
> > > the extended distance with the new super fiber for longer
> > > distances at added
> > > material and installation costs.
> > >
> > > We have to device to products to make sure all levels of users are
> > > affordable.  One area I do not quite agree with you is that: quote,
"the
> > > case of 10 Gig, due to its novelty and likely expense, I think we can
> lay
> > > down new standards for fiber to support it, and customers would go
along
> > > with the concept".  There are some users will fit into your
expectation;
> > > however, there are many will take a bite, only if it is truly
> > > cost-effective, but not novelty.
> > >
> > > We do not have to worry about DMD, or very low-BW cables, which you
> agree.
> > > When users has problems with bad cables, they will find out by
excessive
> > > errors or re-tries.  They will call field servicemen to fix for
> > > them.  Those
> > > servicemen will understand how to identify the bad cables to be
> > > removed from
> > > service.  All we have to do is to educate field servicemen how to
> identify
> > > them.
> > >
> > >
> > >
> > > -----Original Message-----
> > > From: Michael M. Salzman [mailto:msalzman@ibm.net]
> > > Sent: Wednesday, May 19, 1999 5:32 AM
> > > To: stds-802-3-hssg@majordomo.ieee.org
> > > Subject: RE: IEEE 802.3 Requirements
> > >
> > >
> > >
> > > Hi Bruce, Ed, et al,
> > >
> > > My experience is that on average no installer tests fiber for any
thing
> > > other than connectivity.  To imagine that these fellows will carefully
> > > follow some timeconsuming and deliberate procedure to check out
> > > the cable is
> > > wishful thinking.  You are right that DMD is an innate 'feature' of
most
> > > installed base fibers.  The ethernet community only recently
> > > discovered its
> > > ubiquity and impact.  Nevertheless we can expect the vast
> > > unwashed masses to
> > > blithely ignore any special handling and testing requirements untill
> their
> > > installation fails to come up and then they will search for causes.
> > >
> > > In the case of 10 Gig, due to its novelty and likely expense, I
> > > think we can
> > > lay down new standards for fiber to support it, and customers
> > > would go along
> > > with the concept.  For those stubborn souls who insist on deploying
the
> > > technology on existing fiber, I think the 10GEA can develop
> > > recommended test
> > > procedures and recommended equipment sets, and perhaps work with BICSI
> to
> > > develop a "gigabit technician" certification program.   If we do
> > > not address
> > > this prosaic aspect of our industry we are doomed to seeing more trial
> and
> > > error installations.
> > >
> > > Mike.
> > >
> > > > -----Original Message-----
> > > > From: owner-stds-802-3-hssg@majordomo.ieee.org
> > > >  Behalf Of Chang,
> > > > Edward S
> > > > Sent: Monday, May 17, 1999 06:36
> > > > To: Bruce_Tolley@3com.com; Chang, Edward S
> > > > Cc: Thomas Dineen; stds-802-3-hssg@majordomo.ieee.org
> > > > Subject: RE: IEEE 802.3 Requirements
> > > >
> > > >
> > > >
> > > > Bruce:
> > > >
> > > > The 62.5 um fibers have been used for over 15 or 20 years.
> > > > However, we just
> > > > started paying attention to the DMD fibers two years ago.
> > > Because of the
> > > > small DMD population, industry did not paying attention to it.
> > > > In fact, the
> > > > DMD fibers have been with us since the graded index fibers were
> > > > introduced.
> > > > DMD fibers are defected parts like any other products having
> > > > defected parts.
> > > >
> > > > In the past, the MM fibers, particularly 62.5 fibers, were only
> > > used with
> > > > LED sources over-filing the fibers, which is hardly capable of
> > > creating a
> > > > DMD results except low bandwidth.  It was until GbE requiring laser
> over
> > > > 62.5 um fiber, the DMD has not been identified as a detrimental
> > > > problem.
> > > >
> > > > The fiber vendors do not characterize "DMD" as one of the parameters
> in
> > > > their commercial specification due to its very small, negligible
> > > > population.
> > > > As a result, no one really pays attention to verify it, unless an
> > > > experienced engineer is purposely looking for DMD fibers.  Some
> > > time, just
> > > > to identify DMD fibers from the waveforms is not a simple job.
> > > >
> > > > Ed Chang
> > > > Unisys Corporation
> > > > Edward.Chang@Unisys.com
> > > >
> > > >
> > > > ----Original Message-----
> > > > From: Bruce_Tolley@3com.com [mailto:Bruce_Tolley@3com.com]
> > > > Sent: Friday, May 14, 1999 5:29 PM
> > > > To: Chang, Edward S
> > > > Cc: Thomas Dineen; stds-802-3-hssg@majordomo.ieee.org
> > > > Subject: RE: IEEE 802.3 Requirements
> > > >
> > > >
> > > >
> > > >
> > > >
> > > > Ed:
> > > > As far as I know, hardly any "DMD fibers" have been identified to
> date.
> > > >
> > > > Bruce
> > > >
> > > >
> > >
> >