RE: Short haul PMDs
In our experience, the CX was more or less non-existent. It would have been
just fine for the data centers, but only if the parts were available. We
could not acquire most of the various components from our vendors at all.
They were telling us it was due to 1000BASE-T being *the real* copper
standard and they were all just waiting for it to be finalized. 1000BASE-T
is just now shipping (this summer) from most vendors and so I suspect the
market will finally begin to really grow.
As to whether or not it will make up for a lot of our existing in-the-wall
connections, I think it will not. The number of places, outside of our data
centers and wiring closets, in the near future, that need more than
100BASE-T FDX capacity are very few. Of the 1300 machines in one of our
larger buildings, 1200 have only a single 33Mhz/32-bit PCI bus with a
realistic bandwidth of maybe 60-80% of GigE's one-way transfer. However,
their real-world practical LAN capacity is probably bursting around
200Mbits, *if* they have some special application that requires large,
direct to memory-only transfers. (They don't.)
1000BASE-T is mostly replace multiple our aggregated 100BASE-T connections
and provides for lower than a 1000BASE-SX price point. The lower costs let
us avoid QOS or load-shared 100BASE-T implementations when we can fix it
with more bandwidth. I think the same logic will affect 10GE low-cost PMD's
as well, with the primary difference being there is not a low cost 10GE PMD
currently being proposed. (my perception only maybe?) If this view is
wrong, please let me know. I suppose my feeling is that 850nm MMF serial
seems cheaper for short distance links, just like SX is relative to LX, and
T is to SX. Mind you that this is only an emotional feeling based on past
experience and in no way reflects on the real costs to produce these things.
That is for all the fine real engineers to figure out.
In this much I agree. If a better-faster-cheaper short-haul interface is
promised, but not yet available, as far out as we are now from real 10GE
hardware, it may certainly serve to squelch other viable alternatives.
From: Geoff Thompson [mailto:gthompso@xxxxxxxxxxxxxxxxxx]
Sent: Monday, August 07, 2000 2:21 PM
Cc: marshall@xxxxxxxxxxxxxx; stds-802-3-hssg@xxxxxxxx
Subject: RE: Short haul PMDs
I would add to your point that much of the anecdotal conversation that I
have actaullay heard over the years since "z" has been that to NOT use CX
for short haul but rather use 1000BASE-T in that application space on the
assumption that 1000BASE-T was going to be a "volume" product where CX
presumably was not. I always assumed that folks were believing that
1000BASE-T would make significant volume in the 100m in-the-wall
application space. That is an assertion that has not been realized to date.
The cross effect of CX on 1000BASE-T and vice versa is something that
should be included in the conversation when/if we reopen the objectives for
short-haul and talk about broad market potential and economic feasibility.