Re: XAUI, SFF connectors
Early 10 GbE work even predating the March 1999 Call for Interest
focused on looking at solutions for carrying 10 Gbps of data across PCB
traces as well as the media. 10 Gbps serial was difficult and expensive
for both. Other techniques such as parallel lanes and multi-level
signaling were proposed as early as the Call for Interest. Many
proposals focused on 4-lane solutions, bringing the data rate
requirements down to 2.5 Gbps per lane. This was deemed reasonable for
the timing of 10 Gbps product availability.
A group of folks from 802.3, T11 (Fibre Channel) and InfiniBand met
starting in the summer of 1999 to further develop 4-lane logic
architecture which could meet the needs of PCB trace transport as well
as skew, initialization, etc. requirements of 4-lane physical variants.
One strong desire was to do it the same way for all three
standard/industry efforts. By November, 1999, the Hari group had general
agreement that its goals could be met and the contributing individuals
went back to their own organizations to promote a common Hari.
The 802.3 HSSG later renamed Hari to XAUI.
So the real answer to your question is (d). 10 Gbps is too
difficult/costly with current technology, 2.5 Gbps is workable. (a) and
(b) are benefits, not objectives of XAUI.
Chris Simoneaux wrote:
> Maybe someone can help me understand this....
> Was XAUI initially driven by:
> a. IC vendors as an effort to reduce the number of leads on the
> b. PC board vendors to reduce the number, and complexity of traces
> on a board
> c. OEMs for the reasons a. and/or b.
> d. Other?
> -----Original Message-----
> From: Joel Goergen [mailto:joel@xxxxxxxxxxxxxxxxxxx]
> Sent: Friday, July 28, 2000 9:26 AM
> To: stds-802-3-hssg@xxxxxxxx
> Subject: Re: XAUI, SFF connectors
> I agree with your comments stated below. Though I still don't believe 8b10b
> the best choice for XUAI, I strongly believe XUAI is necessary and I support
> completly. Thanks for taking the time to get the points across.
> Take care
> "Simon L. Sabato" wrote:
> > Roy,
> > I'm quite happy to have go through the "growing pains". If this means
> > beating FC and IB to 10G rates, that means that some of their applications
> > will migrate to Ethernet. As they climb on the XAUI bandwagon, it will
> > serve to increase the market for many of the components (retimers, etc)
> > go into 10G Ethernet boxes, as well as increasing the amount of expertise
> > the core technologies.
> > Imagine a world where there's a limited number of experts who can put
> > together a high speed interface such as XAUI (this shouldn't be very
> > just look around you). Now, imagine two possible scenarios. The first in
> > which 10GE, FC, and IB all have different physical layers, and the limited
> > number of experts is split up into three groups. The second in which all
> > have very similar physical layers and the limited number of experts work
> > parts that could be used in each application. Now, you tell me, which one
> > of the above scenarios gives the end customer the lowest cost, highest
> > quality product?
> > You seem to assume that if FC and IB reuse our technology then we are
> > "paying for this and the other technologies are the actual beneficaries".
> > You're partially right, we are paying for the development -- that's the
> > price of being on the cutting edge. Trailing technologies will always
> > leverage the latest technology.
> > But your continuous assertions imply that we are paying for development of
> > XAUI features that *are not useful in 10GE*. These assertions need to
> > or backed up with some information that makes sense. I have yet to see
> > information. I've seen a lot of explanations from varied people at
> > companies explaining the benefits of XAUI. I find it odd that you can be
> > sure that *no-one* needs something. It's easy to know what *you* need,
> > how can you speak for everyone else?
> > In summary, I see XAUI as a way to use a common technology across various
> > applications due to the shared market and shared expertise. Using TTL
> > voltage levels across various applications is a *good thing* even if, say,
> > it wasn't particularly ideal for each one. The only argument against this
> > is that we are saddling XAUI with burdens from other standards, which you
> > claim but cannot substantiate.
> > -Simon Sabato
> > -Manager, Product Architecture
> > -InterNetworking Operation, Intel Corp.
> > P.S. Speaking of what is "costing" the 10GE group, constant discussion on
> > subjects that are closed costs us all. It just cost me twenty minutes to
> > write this message, and a whole lot to read all the others.
Richard Taborek Sr. Phone: 408-845-6102
Chief Technology Officer Cell: 408-832-3957
nSerial Corporation Fax: 408-845-6114
2500-5 Augustine Dr. mailto:rtaborek@xxxxxxxxxxx
Santa Clara, CA 95054 http://www.nSerial.com