Get to the bottom of 10 Gig
you going to deploy a 10 GB Ethernet network? Here's what you need to know to
make an informed and intelligent decision. by K K Shetty
Over the past few years, new applications such as high-resolution graphics
and complex scientific modelling continue to pressure bandwidth requirements
at the desktop and backbone. High-bandwidth applications and high-speed communications
between networked users have driven major increases in desktop computing power,
first to Fast Ethernet and now to Gigabit Ethernet.
This trend is expected to continue with Fast Ethernet becoming the default speed
on the desktop. With rapid advances in micro-electronics and optical networking
technologies that are driving down the complexity and the cost, many customers
have turned to Gigabit Ethernet (1,000 Mbps) to maintain acceptable LAN performance
levels. In less than ten years, data rates implemented on horizontal cabling
systems have grown from 10 Mbps to 100 Mbps and to 1,000 Mbps (or 1 Gbps) today.
Fast Ethernet has become the default speed at the desktop and 1 Gb is commonplace
in the riser and backbones. Servers are currently shipped with 10/100/100 NICs.
It is interesting to note that todays servers will be tomorrows
desktops. One Gb pipes in the horizontal require more than a 1 Gb pipe in the
backbone and data centreshence the drive for 10 GB application standards.
In an ideal world, the migration to 10 Gigabit would be seamless, meaning 10
GbE would run on existing cabling infrastructures up to a distance of 100 metres
on balanced twisted-pair and up to 300 metres on multimode optical fibre. Standards
already enable fibre solutions to 300 metres, but the 100-metre distance on
twisted-pair media has been difficult to define.
Systems backed by existing industry standards offer designers, installers and
end-users value and peace of mind by ensuring inter-operability, performance
guarantees and network expansion. Standards-based solutions limit the possibility
of becoming locked into a proprietary solution that may not survive over time.
So what is the status of 10GbE standards?
Although there were several proposals to enable a copper-based PHY, the IEEE
initially published a 10 GbE Standard recognising only optical fibre, including
the new 50-micron MM Fibre (850 nm laser-optimised 50/125 micron as per TIA-492AAACC),
as acceptable media.
Almost two years later, a 10GBASE-CX4 Standard (IEEE 802.3ak) was published
enabling 10GbE on two-conductor coaxial cable, but only to a distance of 15
metres. The low cabling cost and the low-cost LAN electronics on UTP are stimulating
the market growth for 10G on UTP, hence IEEE is now working on a new project
that hopes to define a 10GbE Standard on unshielded twisted-pair cabling (IEEE
Of the available options for 10GbE, optical fibre is the only standards-based
LAN solution at this time, with cabling and electronics available since three
years. Optical fibre therefore has a significant head start on cost reductions
and product development over twisted-pair. Since STP as a medium is already
defined in TIA and ISO, it is ahead of the standards game compared to the emerging
Augmented Category 6 UTP standards. While STP has been widely deployed and is
a proven solution worldwide, Augmented Category 6 UTP is still in its infancy,
with industry standards still pending, and performance and inter-operability
parameters not yet defined.
The number one concern among designers, installers, and end-users
is system performance. Simply put, the higher symbol rate of 10G requires higher
signal bandwidth. Increasing the bandwidth alone does not resolve all the issues.
Of the three 10GbE options, optical fibre offers the greatest transmission distance
and most bandwidth potential. Todays multimode optical fibre supports
two variants of 10 GbE (10GBASE-SR and 10GBASE-LX4) and Fibre Channel to 300
Just as most installations today use optical fibre in
the backbone and twisted-pair in the horizontal, it may make sense to
combine two or more media types such as fibre, STP and UTP
Fibre cable is also lighter and smaller in diameter than copper
cable, and offers complete immunity to noise and better security from unauthorised
tapping. Compensation for cross-talk and echo can be addressed through brute-force
digital signal processing, but at these data rates and using the faster encoding
schemes, the radiated electrical noise (cross-talk) from other cables becomes
problematic. This alien cross-talk (Alien NEXT or ANEXT) cannot
be sampled, nor can it be measured, so it is not easily compensated, unlike
internal cross-talk such as near-end and far-end cross-talk. ANEXT is agreed
to be the limiting noise source, particularly since 10GBASE-T is self-disturbing
(noise from other 10G links adversely affects each 10G link).
Optical fibre has long been considered too expensive. However, installation
and component costs have decreased dramatically over the past decade due to
longer distance ratings, strength, innovations in connector technology, network
architectures and equipment, and increased familiarity. Today, the cost of a
complete, centralised optical fibre system can be as little as 5 percent more
than a traditional Category 6 UTP installation. Optical fibre cabling offers
the maximum port utilisation and highest density, resulting in overall cost
Historically, however, the industry has preferred copper solutions. In India
UTP solutions dominate STP solutions, primarily due to cost. Other considerations
such as installation hours, density, and grounding/bonding issues have also
had an influence. STP product cost and installation time carry a premium over
todays UTP equivalents (i.e. Category 6 UTP), but not necessarily over
Augmented Category 6 UTP solutions hitting the market.
The historical view that UTP is cheaper may no longer apply to 10
Gigabit. While the industry has moved to adopting smaller and higher-density
solutions to save costly real-estate, Augmented Category 6 UTP is moving towards
a solution comprising larger cable and lower port density. To reduce ANEXT,
Augmented Category 6 solutions have been designed with a larger overall cable
diameter to move the pairs further away from the cable jacket, resulting in
a stiffer and more expensive cable, and in decreased density in cable trays,
conduit, cable management and cable routing.
Since the Augmented Category 6 UTP standard is not yet defined, implementing
these systems today also carries the risk of not meeting standards in the future
or becoming locked into a proprietary solution, possibly resulting in costly
future maintenance. This risk factor combined with density premiums can end
up making an Augmented Category 6 UTP channel costlier than an STP channel.
Unlike fibre, however, STP and UTP solutions utilise the same electronics, so
there is no premium for network equipment.
If migration to 10G is a consideration for a network installation, options are
available in all media today. Optical fibre and STP provide a proven and cost-effective
solution that will support standards-based structured cabling networks. The
UTP announced recently by Tyco for 10G has the capacity to offer the necessary
marginal performance as standards evolve. Though solutions are now available
in both fibre and copper, the picture is less clear from IEEE 802.3an for UTP
Therefore, when considering the network cabling options that are available,
it may help to consider each segment of the network individually. Just as most
installations today use optical fibre in the backbone and twisted-pair in the
horizontal, it may make sense to combine two or more media types (fibre, STP
and UTP) rather than make the decision on an all-for-one basis.
The author is Country Manager, AMP NETCONNECT, India &