Your
server is working slowly and the accounts department is
complaining that queries are taking three times as long as normal.
You start by checking everything from the top of the network stack
down.
The applications seem to be performing alright, and there are no
viruses or unexpected packets in the system. There are no
Trojans on the server, and no one is downloading gigabytes of
music to their
iPod. You are stumped. But did you think of going to the
communications room and checking the server's patch cable?
It is easy to forget that cables are the arterial systems of our
networks. They can get clogged by too much data, and occasionally
they can become degraded or corroded, forcing network managers to
perform the equivalent of a triple bypass to get things running
smoothly again.
Maintaining a healthy cabling system is crucial to the
performance of the network, but how do you measure the health of
your infrastructure?
Cabling: the basic structures
There are three main parts to most cabling structures: the core
backbone cabling infrastructure is usually a vertical cabling
system (running between floors) the structured cabling runs to the
desk (normally horizontally along each floor) and the third part,
the patch cabinet, is where computers on the local area network are
connected to each other and hooked to incoming wide area network
connections.
Patch cabinets are where things can go wrong, because cables are
being pulled and replaced all the time.
Kelvyn Dale, managing director at network consultancy and
cabling firm
Q2 Communications,
says, "When there is a problem, it tends to be the patch cords. The
cables that you see in buildings that get pulled through risers,
and so on, are solid, single-core copper, but patch cords are
stranded, with strands of copper in each of the wires in the cord,
so they tend to break more easily."
But just because it stays put when installed, backbone and
structured cabling is not invulnerable. Even external cabling
designed to run between buildings has unique vulnerabilities. Some
suppliers have introduced anti-rat cables to stop data-hungry
rodents doing their worst, for example. And Dale has seen other
cabling that has given out after sitting in water-flooded ducts for
months on end.
"We often find that electrical contractors pull the cables
around as they would with electrical cable, and that is wrong," he
says.
Cables have a certain bend radius beyond which they are not
meant to flex. To see why, bend a thin piece of copper backwards
and forwards and see what happens to it.
Cables get trodden on and bent backwards, all of which can
degrade the information-carrying capabilities of the wire. Even
having cable ties tied too tightly can cause problems.
Mapping and tracking cabling
When the experts do it, they do it properly. Mobile network
operator
T-Mobile is working with supplier
Computacenter Services to lay new cable at T-Mobile's corporate
campus and 14 switching centres - and there is a lot of cable.
The campus roll-out covers 48,000 outlets on the site and
involves enough cabling to stretch from London to Moscow.
Computacenter is currently working on upgrading T-Mobile's 14
switching offices, which handle voice and data services to
customers.
"We have a complete strategy document that covers the way that
cables are laid in buildings," says Tony Alger, manager for central
and data products at T-Mobile. Alger says that such missives come
from the engineering and planning department. "We have colour codes
detailing what carries what service."
Cables carrying customer data, for example, use a different
colour from those carrying support data or telephone voice call
data. "If you drop something down into the void and it breaks
through the protective coating and hits the cable, then, depending
on what colour it is, you know how important it is to get things
fixed," Alger says.
The void is the area of the datacentre where structured cabling
runs and rarely gets changed. Experts carry out best practice
everywhere, including here, but some amateur installers do not
bother colour coding or documenting patch cabinets, where the
environment changes frequently. That can quickly turn it into a
cable installer's worst nightmare.
"When you get into the cabinet and you find the 'bowl of
spaghetti' syndrome, that takes a long time to do current state
analysis," says Computacenter's IT facilities solution unit
director Neil Silverstein.
Sometimes, network teams will simply add another cord of any
colour for expediency when carrying out a change on the network,
and before long they will begin looking for another cabinet because
their current one is full.
Often, they could recover 50% of a cabinet's capacity by pulling
redundant cables, says Silverstein - if only they knew which was
which.
Why and how to upgrade cabling
But why do companies such as T-Mobile want to upgrade their
backbone and structured cabling? Depending on the infrastructure
and application requirements, such upgrades can come around more
often than you think.
"The industry-prescribed renewal rate is about six years, but
that varies widely in practice," says Rey Geelani, managing
director at cabling specialist Network Interlinks.
Renewal rates depend partly on the situation of the building. A
building with tenants that change regularly may need more upgrades
to suit their particular needs, for example, and existing cabling
configurations or types of cable may not be adequate.
Suppliers suggest installing cable with enough capacity to
support two to three generations of active electronics.
Which cable?
The types of cable have multiplied over the years. Copper and
optical fibre are the two basic types. Fibre is used mainly for
vertical backbone deployments, and copper is generally used for
structured cabling.
Copper breaks down into multiple categories, and the categories
are numbered. The categories are referred to as "Cat", which gives
us Cat 1
cable, through to Cat 6 and beyond.
"Pre-Cat 5
was all based on telephone wiring.
Cat 3
was a telephone-grade cable and was a cheap and cheerful way of
driving a 10mbps network," says Geelani.
Cat 4
was better quality, but was short-lived. "It did not really do
anything apart from deliver 10mbps a bit better."
Cat 5 was the quantum leap in cabling, because it delivered
100mbps, but this was superseded by Cat 5e, which offered up to
1gbps by using a higher grade of copper and rearranging the twists
of the cable inside the shielding.
When Category 6 came along, it supported higher data rates using
a thicker copper core, and supported data rates up to 10gbps - sort
of.
Pre-Cat 6 wiring was able to support distances up to 90m (plus
10m extra for patch cabling) before it needed a repeater, but Cat
6, when pushed to 10gbps , may only be able to cover half of
that.
"New technologies such as 10Gbase-T do not support Cat 5e, and
Cat 6 only does so to certain distances and with much mitigation to
lessen the effects of alien crosstalk," says Carrie Higbie, global
network applications market manager at cabling manufacturer Siemon.
Alien crosstalk refers to signal interference between cables, which
presents both a reliability and security issue.
Companies must also choose between shielded and unshielded
cable, depending on the electrical interference that they expect
from other cables and mechanical systems in areas such as lift
shafts.
The other problem with Cat 6 cabling is that it can be harder to
run through containment systems, says Dale. "Cables tend to be
about 7mm in diameter for Cat 5e and 9mm-10mm for Cat 6. It does
not sound a great deal, but when you have 100 of them in a
container and you have to get them around a corner, those extra
millimetres are important," he says.
In spite of its disadvantages, Cat 6 is the recommended
requirement in the US-based TIA-942 datacentre cabling standard.
However, experts recommend using the next grade of standards as
they become available.
Cat 6A (augmented) operates at double the frequency of Cat 6,
and
Cat 7
drives frequencies even higher.
Fibre cables are just as fragile as copper. Bending them beyond
their allowed radius can introduce hairline cracks that can impede
the signal, says Dale.
There are two main types of fibre - single mode and multi-mode.
The single-mode fibre, which allows just one light source to pass
along the cable, is typically used outside the datacentre for
covering longer distances. The multi-mode version, which allows
multiple light sources down the cable for increased bandwidth,
covers shorter distances and is ideally suited for the
datacentre.
Fibre is graded using the Optical Multimode (OM) standard, and
experts recommend the highest grade, OM3.
That is the problem with cabling: it is expensive, because you
are constantly advised to buy the latest, most future-proof
materials. And the volatile price of copper of late has driven up
prices across the board, says Dale.
But experts argue that you are better off putting in the best
possible cabling and amortising the cost over a longer period,
rather than having to go back and recable sooner because you used
inferior materials.
Options for building
high-speed networks >>
Fibre optic networks
vulnerable to attack >>
Comment on this article:
computer.weekly@rbi.co.uk