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

Re: Selecting PHY Solutions

Jonathan Thatcher wrote:

> All,
> It appears to me, having followed a number of streams over the last two
> weeks, that there appears to be an unusual amount of chest-beating regarding
> the potentials for different PHY implementations. As best as I can
> determine, most of the proposed solutions (sans parallel), appear to be
> closer to a research phase than they do to production. I have no doubt that
> most supporters will say this is true for the competitive solutions, but not
> theirs! :-)
> Might I suggest that while posturing is something that we always have to put
> up with, at this point in time it is more important for us to focus on the
> criteria by which PHY decisions will be judged. We appear to be burning up a
> lot of MIPS on the wrong issues.
> Might I suggest that there are key issues that will affect what PHYs are
> ultimately selected/supported for which we still have no apparent consensus.
> Some examples:
> o Must we, like with 802.3z, attempt to support the fiber infrastructure as
> it exists (e.g. 62.5 micron fiber at 550 meters) in order to accomplish
> Criteria 1: Broad Market Potential? If so, I am POSITIVE we can find a
> solution. It would be a stretch for me to believe that it will be least
> expensive solution. Will it therefore impact the ability to achieve Criteria
> 5: Economic Feasibility (especially, "Reasonable cost for performance
> expected")?

My opinion is NO. Going out on a limb, the Ethernet UTP folks have rarely
introduce a new Ethernet generation without requiring a new cable plant. The
latest example is 1000BASE-T. Does it require just Cat 5, or Cat 5E, or Cat 6?
Apparently, the finished copper standard says that it's cheaper to recable than
to implement a more expensive PHY.

> o If not, is there a sub-portion that we must support in the existing
> infrastructure? For instance: the horizontal space (100 meters)?

The way I follow the logic is:

1) The existing fiber infrastructure cannot support 550 m @ 10 Gbps, but it can
support 100 m cost effectively
2) New infrastructure or a small portion of the existing infrastructure can
support 550m cost effectively

> o If not, are we implying no support for the current building
> infrastructure? If so, will we propose a new building infrastructure?

NO. You can't do 550 m @ 10 Gbps cost effectively on copper. Same for most of
the existing fiber. OK, that means you need new cable.
The current objectives of the copper ad hoc are 2.5 Gbps on cat 6. Is cat 6
considered the existing infrastructure? If it is, than I'd argue that in
equivalent proportions, SMF and high-bandwidth MMF should be considered the
existing infrastructure.

> o What distances are we going to support outside the building? What is in
> the 802.3z objectives; what is in the 802.3z standard (1000BASE-LX); or what
> we are currently shipping in our boxes (A variety of 1000BASE-LX that
> supports the Fibre Channel 10 km specification but runs at 1.25 GBd)?

My opinion: Objective - 2-3 km; Standard 10-15 km.

> o If we cannot resolve these objectives, can we ignore them?

The standard meets and exceeds the objectives after complete and thorough
analysis of all factors. Isn't this our (802.3) reason for existence? We set
reasonable objectives tied to well established requirements, analyze and discuss
multiple solutions to those objectives in a democratic fashion, and then agree
upon a solution which frequently exceeds the original objectives. Is there
something wrong with the the way the Gigabit Ethernet standard was developed
specifically related to this issue? If anything, we set the goals to high and
ended up delaying the standard due to the DMD problem. My opinion: Set
reasonable goals.

> I would suggest that these are primarily marketing questions. What do the
> end customers need in the short term (next 2 to 5 years) and what will they
> need after that? Is 10 Gig ever be a to-the-desktop application? If so, will
> it be this 802.3 effort or some subsequent effort that meets that need? If
> the marketing people say that the price needs to be 3x the equivalent 802.3z
> price at entry to the market and needs to support all of the existing
> infrastructure, what are we going to tell them?

We have the technology to deliver most of what the market wants. The rest
(longer distance requirements on a low-bandwidth fiber) may cost a bit more.

> While I am pounding on the podium, I will note one other thing: There have
> been a large number of bits burned on the reflector regarding WAN
> requirements. But, as far as I know, we are still lacking a concise list of
> specific requirements.
> Jonathan Thatcher "jonathan@xxxxxxxxxxxxx"


Best Regards,

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