RE: [10GBASE-T] Economic feasibility or Market Potential?
I don't understand your statement: "How many "Gig-to-the-Desktop" boxes will be shipping in '04 and '05 ? Assuming a good percentage of them have CX4 interfaces, that will be a large part of the market potential."
I don't see CX4 as a desktop technology. Desktops generally are more than 15 m (as the cable runs) away from the switch.
Some data center connections are under 15 m. Also, given the choice between current fiber costs and careful equipment placement, some people using clusters will be careful to keep distances short enough that CX4 can serve.
However, many distances in data centers are much more than 15 m. CX4 won't answer this need. Also, the cable that carries CX4 for its maximum distance is fairly hefty and not that cheap. I expect that there would be cases especially when at the upper range of CX4 distance where 10GBASE-T would preferred (once it exists) because of the cable management issues of 8-pair twinax cable vs Cat 6 or Cat 7 even if the transceiver cost is higher.
In the early 90's, some of us were concerned about whether Cat 5 cable would become broadly enough installed for a technology running only on it to have broad market potential. History has shown that concern to be misplaced. If at some point, 10 Gig to general desktops becomes important, people will probably pull Cat 7 (or screened Cat 6 if that does it) to be able to run it.
Regardless of what happens in building wiring, there is plenty of space that is 10 m to 100 m in data centers for 10GBASE-T (assuming they can make the power low enough that it can be used in server cards). CX-4 isn't running on installed cable in those data centers either.
We all would love to have 10 Gig running over Cat 5 to 100 m, but it doesn't look like that is going to happen. There is a broad market in data centers for a more achievable goal.
From: DOVE,DANIEL J (HP-Roseville,ex1) [mailto:firstname.lastname@example.org]
Sent: Monday, August 04, 2003 9:01 AM
Subject: RE: [10GBASE-T] Economic feasibility or Market Potential?
> >GE wrote:
> > By the way, I didn't see the same criteria of broad market
> potential applied to CX4.
> > Not that it bothers me;
> Well, it bothers me.
> The more I am following this debate, the more I am becoming
> convinced that 10GBase-CX4
> has no market potential.
Part of the reason that CX4 will enjoy "Broad Market Potential" is that its
cost factor is negligible relative to the available alternatives and it is
based upon technology that is available today. Initial silicon will have a
substantially lower cost than the fiber-optic alternatives, and it can be
quickly integrated into MAC ASSP silicon which makes it even more cost
effective. At ~1W/port, the power impact on device packaging and system
cooling is acceptable, in fact, pretty much equivalent to 1000BASE-T.
How many "Gig-to-the-Desktop" boxes will be shipping in '04 and '05 ?
Assuming a good percentage of them have CX4 interfaces, that will be a large
part of the market potential. I don't think every project has to have
100BASE-T volumes to be considered acceptable, but they must be able to
stand against alternatives.
10GBASE-T has some special challenges, but they are typical of a leading
It must prove feasibility on a medium that will provide it broad potential,
and in a fabrication technology that will provide it low cost. I am not
convinced that CAT7 is going to be broadly enough installed to allow
10GBASE-T to overwhelm its lower-cost alternative CX4 in the data center and
wiring-closet interconnect space. It think that it would be great if
10GBASE-T can run over CAT6 and CAT5e because that will allow it a future
path to the desktop which will drive volume and cost in the right
directions. As a DC/WC interconnect only, it would have to compete with
mature CX4 on the short hauls, and 850nm optics on the longer hauls. That
would really squeeze its potential.
So rather than assume 10GBASE-T will only be a DC/WC interconnect solution
vying against integrated CX4 on one end, and 850nm optics on the other, I
suggest working out a technical solution that supports installed cable
plants (in 05'/06' timeframes) and takes that broader market. If it can do
that, the DC/WC market will open up to it because of its potential
application in the broader space.
HP ProCurve Networking