As IT managers deploy new applications in the data center, bandwidth demand
is outstripping capacity. Applications such as video streaming and the relentless
demand for storage are pushing IT networking infrastructures beyond their limits.
At the same time, IT managers are under pressure to reduce power consumption in the data center as well as the
total cost of ownership (TCO).
Fortunately, the next evolution of Ethernet
is approaching deployment readiness.
The Dell'Oro Group, a market research
company based in Redwood City, Calif.,
projects 10-Gbit Ethernet (10GbE) over
copper will grow to 42% of an 8.8M unit
(total 10GbE) by 2010.
Transporting 10GbE over twisted-pair
copper is by no means a trivial feat, especially
when attempting to provide adequate
signal integrity at the traditional 100-m cable
reach. Moreover, using currently available
process technologies, long-reach transceivers typically
consume more than 10 W.
This will significantly impede the adoption of 100-m 10GbE
transceivers since power consumption above 5 W severely limits
practical deployment of 10GBaseT solutions in the data
center. What is needed is a tradeoff between power consumption
and cable reach.
Making the tradeoff
According to IEEE estimates,
greater than 80% of data-center cable runs are less than 30 m.
For this reason, the recently ratified IEEE 802.3an (10GBaseT)
standard offers a low-power mode (30-m reach), which provides
the ability to trade reach for power. This tradeoff significantly
reduces heat density, power costs, and the capital
expenditures required to dissipate the additional heat.
When viewed from either the system interface or the front
panel RJ-45 connector, it is impossible to identify any differences
electrically, mechanically, or in any other way between a
30-m transceiver and a 100-m transceiver. The only difference
is the length of cable supported and a bit in a management
register that indicates short-reach operation.
IEEE 802.3an clearly defines and completely specifies all
aspects of short-reach operation. It also provides a means to
verify interoperability and compliance to the short-reach
requirements of 10GBaseT by specifying test channels for both
Category 6a (Cat6a) and Category 7 (Cat7) cabling. A transceiver
that meets all of these requirements is in compliance with
the short-reach provisions of IEEE 802.3an.
Signal integrity is another factor that poses a
unique challenge for 10GBaseT transceivers.
It's hard to ignore the fact that at 10
Gbits/s over 100-m copper cable lengths,
we have exceeded the theoretical and
practical signal-integrity limits (i.e., Shannon's
limit) of Category 5e (Cat5e)
cabling. This fact precipitated the development
of a new cable type-Cat6a.
Add to this the need to provide
sufficient headroom (usually
20% to 30% additional cable
length) to satisfy the OEM's rigid
requirements, and the technological
challenges loom larger than ever for 100-m
10GBaseT solutions.
Bigger isn't better
As one would expect in any semiconductor
solution, size matters. The footprint of a solution dictates
cost, port density, power consumption, heat dissipation,
board space, and packaging. Low-power, short-reach 10GbE
transceivers will consume less than 5 W in a very small (singlechip)
footprint, qualifying them for rapid adoption and deployment
either as fixed-port (chip-on-board) solutions or in MSAcompatible
modules (e.g., X2, XFP and SFP+).
Besides providing robust signal integrity and consuming half
the power of present-day long-reach alternatives, deploying
short-reach transceivers in these modular solutions will also
provide OEMs the flexibility to offer a backward-compatible,
legacy-friendly 10GbE-over-copper upgrade path. It also accelerates
time-to-market for the physical-layer (PHY) vendor by creating
an intermediate step toward the PHY's ultimate chip-onboard
deployment in OEM systems.
As power density in the data center continues its inexorable
climb to the top of the IT manager's list of hot buttons, the value
of the short-reach 10GBaseT solution may well become the
pivotal factor in the successful deployment and commercial
viability of 10GbE over copper.