DATA CENTRE

NETWORK CABLING
BEST PRACTICES TO MEET THE
DEMANDS OF THE DIGITAL ECONOMY

SPONSORED BY
 DATA CENTRE
NETWORK CABLING

CONTENTS

4 INTRODUCTION
5 WHY BEST PRACTICE?
7 THE IMPORTANCE OF
REDUCING CABLING RISK
8 DATA CENTRE NETWORK CABLING BEST
PRACTICE THE KEY ACTIVITIES
9 THE FIVE STEPS
9 PLANNING
15 MAPPING
19 PROCUREMENT
25 INSTALLATION
28 MANAGEMENT
30 WHAT TO DO NEXT
32 BIBLIOGRAPHY

2
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

FIGURE INDEX

Page 5 Page 21
Figure 1: Perception of How Inadequate Networking Will Figure 11:  IEEE Technical Specifications for Copper Twisted
Impact Data Centres (2015 to 2016) Pair Cabling
Page 6 Page 22
Figure 2: The Deming Quality Improvement Cycle Stages Figure 12: Technical Specifications of Standards for
Multimode and Single Mode Optical Fibre Cabling
Page 9
Figure 3: How the Network Row Can Be Positioned Closer Page 23
to Server Rows Figure 13: How Data Centre Building Spaces Differ in their
Network Cabling Requirements
Page 11
Figure 4: The Different Spaces of a Data Centre Building Page 24
Figure 14: Reach and Speed Specifications of Copper and
Page 12 Fibre Network Cabling
Figure 5: The Cabling Categories That Cover A Data Center
Figure 15: Copper Network Cabling Usage in Different Local
Page 14 Distribution Connections
Figure 6: A Data Centres Network Cabling Requirement
Check List Page 26
Figure 16: A Typical Ultra-High Density Fibre Panel
Page 16
Figure 7: The Three-Tier Network Topology Page 28
Figure 17: Network Cabling Used to Enable a Range of
Page 17 Services
Figure 8: A Network Topology Updated for East-West Data
Centre Traffic Page 29
Figure 18: Examples of Secure Lock to be used on Copper or
Page 18 Fibre Patch Cords
Figure 9: Different Connection Configurations at the Local
Distribution Level Page 31
Figure 19: Nexans Delivery Capabilities
Page 20
Figure 10: The CENELEC EN 50600 Telecom Cabling Series
Standards

3
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

INTRODUCTION

As the importance of data centres grows, so too does the need for high-quality networks in
order for the demands of the digital economy to be met. The modern data centre relies on
quick and efficient exchanges of information, to and from customers and between the active
IT components it houses. As such, growth in the deployment of virtualization and of cloud
systems within the data centre mean that low latency, adequate bandwidth, connections and
ports are as important as the adequate provision of power and cooling.

The need for proper decision making and management in relation to network equipment,
including cabling, is driven by data centre operational requirements. As with the
equipment it connects, cabling should be modular, scalable and easily adaptable. It is able
to transmit more data, faster and more accurately without an equivalent increase in the
power or space required to do so. Improved design and management practices mean
less time and labour are spent maintaining, repairing or installing cabling. More than ever,
planning, installing and maintaining network cabling to standards of best practice gives
operators a greater chance of chaos avoidance in the data centre.

Data centres typically contain many kilometres of network cabling in addition to thousands
of connections and ports that can be affected by physical stress, electromagnetic noise,
poor installation and connection, heat or humidity. Without knowing fully the location and
role of each length of cabling, its properties, what it is meant to be connected to, and
without some prior planning to replace the faulty length then a temporary interruption to
communication within data centre operations can become a larger problem.

Therefore, the rules of best practice in relation to data centre network cabling will be
based on a number of key steps:

a long-term view of a data centres cabling equipment requirement

rigor in the selection and mapping of the data centres cabling infrastructure, and
installation and management of cabling in a simple, systematic and replicable
manner.

4
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

WHY BEST PRACTICE?

The key purpose of the modern data centre is to house and transmit data between active
IT equipment servers and storage and to and from users outside the data centre. The
digitisation of business and society has created an exponential growth in data traffic with
consequent pressure on data centre network requirements. According to Ciscos Global
Cloud Index (2015), global data centre IP traffic will grow at a CAGR of 25% between 2014
and 2019 and data centre workloads within traditional (as opposed to cloud) data centres
will more than double (1).

The importance of networks to the data centre means a greater management focus on
how well the networks underpinning the data centre will meet its requirements. A majority
of data centre owners and operators in 2015 (58.6%) expressed concern about the
threat that inadequate data centre networks pose to their data centre operations (2). This
positions the network only slightly behind operational costs and the shortage of suitably
skilled staff as a major source of concern for operators:

Figure 1: Perception of How Inadequate Networking Will Impact Data Centres (2015 to 2016)

END-USERS COLOCATION CLOUD TOTAL

Significant threat 23.2% 17.9% 16.2% 21.5%

Some threat 35.6% 37.9% 44.5% 37.1%

Little or no threat 41.2% 44.2% 39.3% 41.4%

5
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

The convergence of servers, storage and network systems in the The principle behind Quality Improvement Cycle (QIC) is to review
data centre further creates an impetus for scalability and flexibility your data centre networks to plan how they can be improved, and,
in cabling deployment. The growth of hybrid cloud systems and as part of that, to define the requirements/targets that need to be
virtualisation in meeting corporate met. The do stage will map the
IT requirements represents another network design to the requirements
pressure on data centre network THE FAILURE TO FOLLOW BEST and configuration of the data
cabling since, at 2015, the DCD PRACTICE CAN ALSO COST MONEY centre and to the network solutions
Intelligence Census indicated that THROUGH THE FAILURE TO RECOGNISE available for the task which will then
17% of global footprint is located AND CAPITALISE ON OPPORTUNITIES be installed into the data centre.
in virtualised or on-premise cloud FOR BETTER DATA CENTRE OPERATION. Checking includes the testing of
environments (3). These internal equipment as well as ensuring
systems require more complex, that it continues to meet network
higher specification and lower latency network systems. requirements. Further action may be required to add, replace or
modify equipment to ensure that requirements continue to be met.
These changes will impact how the requirements of data centre
network cabling will need to evolve in the future. Only the As with any data centre component, the case for best practice can be
application of best practice and the availability of new cutting-edge viewed in terms of reducing the risk to operation and performance,
network cabling solutions will enable the modern data centre to and as the opportunity to develop improvements to both:
deliver on its networking requirements.
It reduces the risk of downtime/interruption to availability
Any best practice model such as the Deming Cycle needs to and the associated costs of this income if a data centre is
be continuous when used to support the availability, efficiency used for selling or associated revenue generation activity, the
and performance requirements of a data centre. Even the best damage to reputation, the damage also to company morale
designed network system will fail if it is not properly installed, and the costs of restoration
maintained or managed. The process starts pre-commissioning and The failure to follow best practice can also cost money
ends only once de-commissioning is finished. through the failure to recognise and capitalise on
opportunities for better data centre operation through
reduced operating costs, better use of space, or the reduced
Figure 2: The Deming Quality Improvement Cycle Stages need to build or extend the data centre.

PLAN

CONTINUAL
ACT DO
IMPROVEMENT

CHECK

6
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

THE IMPORTANCE
OF REDUCING
CABLING RISK

The analysis and reporting of major data centre outages indicate that network cabling
is not often the cause of a major data centre outage or shutdown. Data centre outages
or shutdowns are typically a result of power distribution, power protection failures or of
thermal events caused by IT overload and/or malfunctioning cooling systems. However,
failures of network cabling equipment can have a negative impact, as demonstrated in
2013 by reports of a poorly designed switch that, when inserted, hit the reset button on
the back of the equipment, wiping data when plugged in. The Uptime Institute notes,
incidents that may threaten the operation of the data
centre do not always result in outages or shutdowns (4).
However, a number of sources suggest that network
cabling accounts for a high proportion of such incidents.
CABLING RUNS THROUGHOUT THE
Emerson/Ponemon Institute studies quoted by the Uptime DATA CENTRE, WHICH MEANS IF IT IS
Institute (5) indicate that incidents occur daily which POORLY INSTALLED OR MANAGED, IT
means that if the proportion of incidents indicated by DCD CAN BE A NUISANCE OR AN IMPEDIMENT
Intelligence research is followed, the identification and TO PROGRESS.
correction of network cabling system incidents happens
at least weekly in most data centres (6).

Cabling runs throughout the data centre, which means if it is poorly installed or managed,
it can be a nuisance or an impediment to progress. If there is network failure in the data
centre and it is not possible to quickly identify and access what needs to be replaced,
then additional time and effort will be required to do this. If the cabling is not readily
accessible because of poor installation then adjacent sections of the data centre
will need to be shut down while staff or contractors go hunting for the rogue link or
connection. In this situation it should be remembered that a considerable proportion of
unplanned downtime can be linked to planned downtime that overruns or creates new
problems in the data centre. Using a process of trial and error to identify the culpable link
or connection has the potential to add to the initial problem.

7
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

DATA CENTRE NETWORK

CABLING BEST PRACTICE
THE KEY ACTIVITIES

The key stages in a best practice program for data centre network cabling are:

1
In order to establish the network requirements of
Planning the data centre in order that the deployment is
consistent with current needs

Ensuring that on-the-ground decisions reflect the

2
objectives established through planning, through
Mapping mapping the requirement onto the actual data
centre using the most suitable network topology
design

Making the correct choice of equipment for the

3 Procurement
network in terms of density, speed, reliability,
efficiency, service and support, return on
investment etc.

4
Since a number of the problems and inefficiencies
Installation associated with cabling in the data centre can be
traced back to poor installation practices

Since any data centre system is only as good

5 Management
as its management and operation, particularly
when it needs to bridge a period of significant
technological change.

8
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

STEP 1 OF 5

PLANNING
THE CRITICALITY OF NETWORK
AND CABLING PLANNING

The planning stage essentially will specify the requirement for cabling and other network
equipment in the data centre. This will then be used in the decision-making processes for
equipment and for the network topography which will provide the blueprint as to how the
equipment is distributed through the data centre.

The planning stage should be adopted across all types of major data centre projects
new build, modular deployments, facility extension and refit/refresh projects. Each of
these different types of data centre investment will present different opportunities and
limitations in terms of network design. A greenfield new build will enable greater latitude
in the design and installation of cabling infrastructure as it will not need to take into
account existing facility or network limitations. New build will also allow the servers and
storage racks to be configured to ensure that they are sufficiently close to network racks
to minimise transmission disruption across the link.

Figure 3: How the Network Row Can Be Positioned Closer to Server Rows

9
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

In contrast, a migration project may simply mean disconnecting groups that design, deploy, plan or oversee IT traffic that passes
the network equipment in one facility and plugging it back in at through data centre networks. The responsibility for network
the new facility. According to DCD Intelligence estimates, the cabling has traditionally fallen between IT and facilities personnel.
majority of networking equipment investment is spent on existing It may also now involve personnel with specific responsibility
data centres (7), most usually in situations where the IT equipment for the network, for cloud or for IT operations. The increased
is in need of an upgrade or where space and/or capacity is being criticality of the network to the data centre should be reflected in
added to the data centre. the greater degree of responsibility taken by all parties to whom
its performance is relevant.
In this context the use of higher-specification
equipment and cabling cannot simply be added This whole of data centre team
into a data centre as it will change the entire THE STRATEGY FOR BEST approach will make it easier also to
network topography of the data centre. If establish and review performance
PRACTICE IN CABLING NEEDS
introducing new network infrastructure then this criteria for the reliability of the
IDEALLY TO BE PART OF THAT
requires a detailed inventory of installations and network and its infrastructure in
what is needed when cabling type, quantities,
WHICH COVERS THE WHOLE addition to (i.e. MTBF (Mean Time
connections, panels, ports, equipment
DATA CENTRE. Between Failure), utilisation, meeting
specification and speed requirements, patching company needs, cost of operation in
and housing are considered. Those involved addition to other factors.
in the planning process need also to consider
for future network cabling requirements network cabling may As network cabling systems are complex and new innovations
have a life of 10 years or more. Therefore the future planning of in design, manufacturing and installation are introduced, many
network cabling is important so that it is able to accommodate companies will enlist network and cabling experts to advise on
the two or more generations of active IT equipment that may be how best to plan their network cabling.
deployed over that time. The introduction of new network units
into an existing data centre will also need to look at any particular
design and installation considerations limitations on space use,
airflow requirements, the use of under floor or overhead supply
systems and current practices in using Top of Rack, Middle of
Rack or End of Row configurations.

It should be noted that while the selection of telecommunications

companies providing the network into which the data centre
feeds is an important consideration, it is not covered here. The
recommendations made here cover the systems within the data
centre and within buildings that house data centres.

The strategy for best practice in cabling needs ideally to be part of

that which covers the whole data centre. If it is not, then it should
draw on that strategy and be shaped by it. If there is no existing
data centre strategy for best practice then the networking systems
may be a good place to start as naturally questions asked about the
networks will bring into play questions about the equipment in the
data centre, and their capacity, efficiency and performance.

WHO SHOULD BE INVOLVED?

In order for network cabling to meet user requirements, it is
recommended that ownership of a best practice strategy involve

10
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING
HOW TO PLAN WELL
What should be the basis of establishing a best practice plan, and
on what basis should the decisions relating to it be made? Each
company will have its own priorities but the underlying need is to
ensure that all components of the network including cables, cords,
switches, connectors, panels, housing, cards, routers, nodes are
configured to connect the companys equipment within the data
centre and to provide links outside the data centre in order that
the companys IT objectives can be met.
Other factors that need to be considered as part of the planning
stage will include:
How the IT capacity of the data centre may need to increase
and also how the IT load may become more variable and
unpredictable due to the demands of convergence, big data
etc. These changes will mean that the network will need
to be flexible and scalable in line with the evolving data
transmission requirements of the data centre.
The deployment of cloud and virtualization. Few data centres
in established markets only house physical racks. Server
virtualization has been a mainstream technology for a number
of years and has been followed by virtualization of users and
application. End-user data centres have also become the
location for private and hybrid cloud systems as companies
prefer cloud options that allow them to keep an eye on their
cloud footprint. The move to software-defined data centres
and networks, however, is relatively new but it will affect
the type of IT equipment found in data centres and the
connection requirements in future. As such, planning needs
to examine how the data centres mix of physical servers,
virtualization and cloud technologies will change.
The cost of the initial investment and of subsequent network
operating costs including repair, renewal, maintenance, the
ease and speed of installation may be important.
Network planning for the data centre needs also to take account
of upgrades and changes in the technological evolution of data
centre network cabling equipment. Since accepted cabling
standards for telecommunications and other commercial purposes
were established 25 years ago, the growth in the carrying
capacity, speed, reliability and ease of integration of network
cabling has been considerable. Other changes have been made
to the physical characteristics of network cabling options. The
strength and flexibility of network cabling have been improved,
11
www.dcd-intelligence.com
cables are narrower, the length at which signal integrity can be
preserved is longer in addition to lower rates of interference from
electromagnetic noise and greater capacity to integrate within
structured systems.
Therefore it is important also to establish the length of time for
which the best practice strategy is designed to be effective
although one of the defining characteristics of best practice is
that regular reviews will identify further needs for change and
improvement.
THE IMPORTANCE OF THINKING
BEYOND THE WHITE SPACE
Most data centres and server rooms are located within buildings
that also house other commercial or industrial activities rather
than in dedicated purpose-built structures of their own. This
means it may be good practice to plan for the whole building
rather than just the data centre or server room space.
Usually, the space within the building can be divided into
three areas:
1 white space which houses the IT equipment
in a data centre or server room
the technical space which houses the electrical,
2
mechanical, gensets and power protection equipment
in addition to the other technical systems the building
requires to function
3 the office or control space which houses staff
and the equipment needed for their work
These spaces may be separated and run independently, or
share common supply infrastructure. There is however a mutual
dependence between them as the computer room/data centre
will require control from the office area as well as support in
terms of power provision and protection (UPS) and mechanical,
plumbing and electrical systems. The requirements for power,
environmental control and security will be different for each space
and this will determine the network profile for each.
 DATA CENTRE
NETWORK CABLING

Figure 4: The Different Spaces of a Data Centre Building
Office Space / Control Room
Office Space / Control Room
The technical space has low density requirements but the need
for its systems to reach throughout the building based on its role
as a monitoring and control centre for supporting the data centre.
The office space requires higher speed connections to connect
its commercial equipment, to enable its telecommunication/VOIP/
PoE systems, wireless networks and cabling between equipment
units. The computer room/data centre requires connections to
Office Space / Control Room

Office Space / Control Room and from the outside world as well as connections between
Office Space / Control Room
active IT server and storage equipment. For switch to switch links
Office Space / Control Room it requires medium reach and ultra-high bit-rate connections; for
switch-to-server and switch-to-storage connections, short reach
Office Space / Control Room
and very high bit-rates are required.
Computer Room Computer Room Technical Space Computer Room Technical Space Technical Space
(White Space) (White Space) (White Space)

Computer Room Technical Space

The different networking requirements of each space can be
(White Space)
coordinated as part of a whole of building approach at the planning
stage, particularly as these different spaces need to communicate
with each other. As more smart and intelligent buildings are
erected, the integration of networks within a multi-purpose building
Computer Room Computer Room Technical Space Technical Space
(White Space)
Computer Room (White Space) (White Space)
Technical Space will become a topic of greater interest and thus more important.

Figure 5: The Cabling Categories That Cover A Data Centre

Fibre cabling
Copper cabling

12
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING
WORKING OUT THE NETWORK REQUIREMENT
It is critical to the process of planning the data centres cabling
requirement that some accurate means of estimating how much
network capacity and therefore how much cabling equipment will
be required. Like the provision of other resources and equipment
into the data centre this is a component of capacity planning
that will look to maintain the optimum position between under-
provisioning where there is insufficient network capacity for the
data centre to do the work it is required to, and over-provisioning
where relatively little of the network infrastructure is used in the
data centres operation.
As with other data centre metrics, the method of calculating the
network infrastructure requirement has become more complex
over time as it reflects the larger number of variables that may
impact on the calculation. The estimate of requirement on the
basis of space is now largely little more than an initial ballpark
step since that was appropriate for an era of largely predictable
and low density IT loads. Some of the factors that need to be
considered when working out requirement are:
The IT capacity of the equipment the network will connect
and how this is projected to change.
Many data centre owners and operators are looking at the
peak requirements that their network will have to carry rather
than an average requirement across the facility. This is a
function of an IT load that is more variable and unpredictable
either as a result of variable customer demand (on-line retail in
13
www.dcd-intelligence.com
December, on-line sports coverage on weekend afternoons,
for example) or due to traffic connecting across a converged
data centre system. This mirrors the growing trend to look
more at the maximum rack density rather than the average
rack density in the data centre. For this reason, it may be
sound strategy to divide the data centre into network zones,
where certain racks of servers have the necessary capacity to
deal with higher demand or to be able to network to virtual or
cloud systems to cope with additional IT requirements.
The requirement for network capacity needs should also
take account of redundancy in the network, although on a
day-to-day basis this will follow the overall redundancy of the
data centre.
Capacity estimation here needs also to factor in the number
of channels running through the network, the type of cabling
and its capacity as well as growth factors based on projected
usage growth.
The requirements for networking equipment is an exercise in
numbers and based on the numbers and the specifications of the
key server and storage equipment that will need to be connected
as well as the network distribution sub-systems that will link the
equipment through and within the data centre. An example of
how such information may be collected and organised is shown in
Figure 6. The key information sets include the numbers of server,
storage and network rows and racks and of distribution areas and
patching frames as well as their height and width, and the number
of connections at these points in the network.
 DATA CENTRE
NETWORK CABLING

Figure 6: A Data Centres Network Cabling Requirement Check List

ITEM DESCRIPTION ITEM DESCRIPTION

Chapter IV A / Data Centre Layout Chapter IV D / Network racks data collection

1 If available please provide the drawings of the DC layout Number of network racks per row maybe positioned at end of server
23
rows
If available please provide the drawings of the cable trays layout
Cable tray configuration 24 Number of rows of network racks
2
Trays located above the racks or under the raised floor? 25 Location of network racks in the computer room floor
Distance from tray to cabinet
26 Height (U numbers) of network racks
3 If available please provide the networking diagram
27 Width and depth of the network racks
4 Type of Data Centre e.g. Corporate, Colocation
28 Switches model and part number
Chapter IV B / Server racks data collection
29 Number and location of the switches
5 Number of server racks per row
30 Number of Ethernet connections to patching frame per rack
6 Number of rows of server racks
31 Number of fibre uplinks to patching frame per rack
7 Location of server racks in the computer room floor
32 Class of copper cabling if specified
8 Height (U numbers) of server racks
33 Grade of fibre cable and connectivity if specified
9 Width and depth of the server racks
Chapter IV E / Distribution area / patching frames
10 Number of copper connections per rack Note: The following data can be calculated by Nexans if all data requested
here above are provided
11 Number of fibre connections per rack
34 Number of copper cabling racks per distribution frame
12 Class of copper cabling if specified e.g. CAT6A, CAT7A
35 Number of fibre channel connections per distribution frame
Grade of fibre cable and connectivity if specified
13 36 Number of distribution frames
e.g. OM3-LC, OM3-SC etc.

Chapter IV C / Storage (SAN) racks data collection 37 Location of the distribution frames in the computer room floor

14 Number of storage racks per row for SAN fabric switches 38 Height (U numbers) of the distribution frames

15 Number of storage rows for SAN fabric switches 39 Width and depth of the distribution frame

16 Location of Storage racks in the computer room floor Number of copper cables to each rack (should correspond with
40
combined number in each server and network rack)
17 Height (U numbers) of storage racks
Number of fibre connections to each rack (should correspond with
41
18 Width and depth of the storage racks combined number in each server and network rack)

19 Number of copper connections per rack 42 Class of copper cabling if specified

20 Number of fibre connections per rack 43 Grade of fibre cable and connectivity if specified

21 Class of copper cabling if specified Chapter V / Added value

22 Grade of fibre cable and connectivity if specified 46 A. LANsense Intelligent Infrastructure Management

47 B1. EMAC in Server racks

48 B2. EMAC in Storage racks

49 B3. EMAC in Network racks

50 B1. EMAC in Server racks

51 C1. Copper Pre-Terminated assemblies

52 C1. Fibre Pre-Terminated assemblies

14
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

STEP 2 OF 5

MAPPING
MAPPING DATA
CENTRE NETWORK REQUIREMENTS

There are two main parts to the purchasing decision process to ensure the network
meets the requirements of the data centre:

MAPPING:
Designing a topography that maps the network architecture from the point(s)
1 at which it enters the data centre building to the point at which it connects with
the server or storage unit and which also allows connection between active IT
equipment.

PROCUREMENT:
2 Deciding on the cabling and networking equipment that will be deployed in the
data centre.

The mapping of the most suitable distribution structure and network architecture will be
based on the following mapping components:

Zones or areas the main distribution cabling sub-system which will be the
central area within the network topography as well as the access point of the

1 system, the intermediate distributor cabling sub-system (where connections

branch off the main distribution sub-system to individual local sub-systems or
groups of these), and the local distributor cabling sub-systems where the network
connects individual work areas, for example storage, server racks or rows.

The switching and network topology (e.g. legacy, fat tree, DCell, BCube,
2 Scafadia) paying particular attention to the switch to switch and switch to
server/switch to storage interfaces

The provision of services to the equipment via Middle of Row (MOR), End of
3 Row (EOR) or Top of Rack (TOR) and the use of structured as opposed to point
to point cabling systems.

15
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

There are a number of factors that may influence the design of
the network architecture as part of broader cabling planning,
including:
Scalability that additional data can be carried across the
network whether on a short-term load-peak basis or as
part of a broader data centre capacity expansion. As noted
previously, networking equipment is so pervasive in data
centres that such scalability needs to be easily enabled.
Designing according to physical constraints of the cabling
link transmission specification, physical flexibility and
strength, minimization of interference in order to increase
robustness and to prevent any major interruption to
transmission
The bandwidth and latency requirements of the equipment
within the data centre and the service levels expected of
it. Thus, the numbers of servers that are virtualised will
increase the complexity of traffic through the data centre
and in particular this will influence the network topology
for switching and routing. The growing emergence of the
data centre fabric which can act as a unified system of
connections between switches and equipment in disparate
locations across the data centre has acted as one solution
to this potential complexity. Yet as with any service that is
THE EVOLUTION OF NETWORK TOPOLOGIES
The topologies that the network architecture has traditionally followed
have been hierarchical, and based around 3 layers of network
switches connecting servers via access switches to aggregate layer
switches which are then connected to each other (north-south) by
core layer switches. These also connect to the Internet. The hierarchy
(or tree) may not be scalable enough to deal with cloud computing
and the design may suffer congestion at its higher levels. Therefore,
this topology has evolved through, for example, its sub-division into
pods to enable more complex switching arrangements, greater
bandwidth and reduced network congestion.
Server-centric building block topologies such as DCell have
been developed to offer greater scalability. There are additionally
topologies based on cube structures, ring-shaped configurations
and more amorphous/free form topologies.
Figure 7: The Three-Tier Network Topology
CORE
Backbone

on demand and this is a key driver for virtual, cloud and

software-defined utilities the network needs to be able to
connect at very low latency and without service interruption.
Distribution
Since the processes associated with virtualization usually
mean the consolidation of numbers of physical servers
into virtual machines then accessed through a smaller
numbers of servers, this increases the speed and bandwidth
Access
requirements for network connections.
Power use. DCD Intelligence has estimated 18 to 20 billion
kWh of power can be attributed globally in 2014 to network
infrastructure or roughly 5% to 6% of total consumption. The
increased operational dependence of the data centre on
Web Server Email Server PC1 PC2
networks will only serve to drive the overall amount up (8).
Costs of installation and of operation.

16
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

Three-layer designs are being execute the data transfer, causing the
superseded by leaf/spine THREE-LAYER DESIGNS ARE potential for congestion there. Leaf/
topologies. In part, this change BEING SUPERSEDED BY LEAF/ spine adds switches at the layer below
is based on the trend in network SPINE TOPOLOGIES. IN PART, THIS the access layer and then spread a
traffic patterns from north-south CHANGE IS BASED ON THE TREND spine of switches across rather than up
(in and out of the data centre IN NETWORK TRAFFIC PATTERNS the network. Leaf/spine is a standard
usually between hosts by the most for Ethernet fabric designs as it creates
FROM NORTH-SOUTH (IN AND OUT
direct route) to east-west traffic less distance and switching for hosts to
OF THE DATA CENTRE USUALLY
which travels across network communicate east-west, greater flexibility
interconnection points within the
BETWEEN HOSTS BY THE MOST in switching protocols than is possible
data centre. In particular this is DIRECT ROUTE) TO EAST-WEST with chassis switching systems. However,
a consequence of convergence TRAFFIC WHICH TRAVELS ACROSS this flexibility requires a higher number of
(servers talking to each other and NETWORK INTERCONNECTION switches and a greater amount of cabling.
to storage units) and the need POINTS WITHIN THE DATA CENTRE.
therefore for traffic between Each of the topologies described
servers/apps and storage, and of briefly here, and others, have positive
virtualization and cloud systems which move workloads via virtual characteristics and drawbacks. Decisions as to which to use will
machines between hosts within segments in the data centre. The depend upon the priorities assigned to the network within the data
increase in crossways east-west traffic within the data centre centre according to corporate IT priorities, and the extent to which
means that it may have to move up to the aggregation layer to the data centre requires scalability and low latency in its operation.

Figure 8: A Network Topology Updated for East-West Data Centre Traffic

INTERCONNECTION INTERCONNECTION
SWITCH SWITCH

SERVER SERVER ACCESS SERVER SERVER

CABINET CABINET SWITCHES CABINET CABINET

Middle of Row (MoR)

ACCESS ACCESS ACCESS ACCESS ACCESS
SWITCH SWITCH SWITCH SWITCH SWITCH

SERVER SERVER SERVER SERVER SERVER ACCESS SERVER SERVER SERVER SERVER
CABINET CABINET CABINET CABINET CABINET SWITCHES CABINET CABINET CABINET CABINET

Top of Rack (ToR)

ACCESS End of Row (EoR)
SWITCHES

SERVER SERVER SERVER SERVER SERVER

CABINET CABINET CABINET CABINET CABINET
Dedicated Network Row

Active switch-to-switch Server connections

connections (optical fibre) (typically copper)

17
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

ACCESS TO THE EQUIPMENT Figure 9: Different Connection Configurations at the Local

Distribution Level
The connection to the active IT equipment can be made via the
top of the rack (ToR), or via the middle of the row (MoR) or via the Top of Rack
end of the row (EoR)

While there has been some move in recent years towards

ToR based on greater convenience of access and improved
space utilisation, a number of recent studies indicate that
energy consumption, flexibility and switch port utilisation may
be improved through EoR connection based on any Base-T
morphology.

IEEE studies published in March and July 2014 (9), and a CCCR
White Paper (10), although based on different cost comparison End of Row
criteria, all indicate cost savings of EOR over TOR connection.
This is based on the further level of connectors required at the
top of the rack for a TOR connection, the need of EOR connection
for fewer port aggregation switches and uplink ports, and savings
on redundant power supply. The structured cabling/EOR option
also achieves a higher level of switch port utilisation. Based on
the assumption of a 39 cabinet data centre at an average of 5kW-
6kW per cabinet and 14 servers per cabinet, the total costs are
worked out as $4,227,000 for ToR access and $2,075,000 for a
structured cabling system (10).
Network Row
Comparatively, ToR is perceived to offer the advantages of fewer
cables located under the floor, cost saving on horizontal copper
cabling and suitability for high density blade servers. EoR also
offers lower latency, a better control over oversubscription and
lower TCO than ToR.

The decision as to how to connect network cabling at the zone/

local distribution levels will depend on the configuration of the
racks and the basis of supply to them.

Network Row

18
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

STEP 3 OF 5

CABLING
PROCUREMENT
DECIDING ON CABLING AND NETWORK EQUIPMENT

The decisions to be made in relation to the choice of the network equipment to be fitted
in the data centre need to consider a number of different factors:

Whether to build a system from disparate components and suppliers or whether

to install a structured cabling system which enables a common warranty on
applications, link and channel usually using equipment from a more limited range
of suppliers
Whether to install fibre, copper or a mix of both
The cabling bandwidth requirements THE DECISIONS
The integration with switching, connectors, TO BE MADE IN
patch cords, patch panels, housing and security RELATION TO
systems, identification systems THE CHOICE OF
Supplier profile their brand and profile, THE NETWORK
their record in innovation and participation in EQUIPMENT TO
developing standards all indicate their expertise BE FITTED IN THE
in and commitment to data centre network
DATA CENTRE NEED
equipment
TO CONSIDER
Cost
A NUMBER OF
DIFFERENT FACTORS.

19
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

THE IMPORTANCE OF STANDARDS
Cabling is an equipment category that has developed subject to a raft
of standards that variously define the performance and capabilities of
both copper and fibre cabling as well as offering guidelines as to best
practice in design, installation and management.
Of particular relevance are the new CENELEC EN 50600 series
standards which will impact on data centres and buildings that
house data centres. The series is separated into practices in
terms of building construction, power distribution, environmental
control, telecommunications cabling and security systems,
as well as the management and operational systems that
oversee all of these. It specifies a classification based around
risks including availability, security and energy efficiency. Four
telecommunication cabling classes have been developed:
Where there is no resilience and high risk such as
Class 1 with direct point to point (unstructured) cabling
Where there is no resilience but fixed cabling
Class 2 systems are deployed
Concurrently maintainable based on multi-path
Class 3 cabling systems
Concurrently maintainable based on diverse routed
Class 4 topologies. This is the lowest risk (but highest cost)
proposition.
The EN 50600 standard is important since it creates a Class
system for redundancy in network cabling. This applies a standard
to the network topology not just to the component products
(cables, switches, connectors) where the standards have largely
been concerned with performance, technical specification and
not risk.

Standards are a core definition of cabling capability which means

they are rarely ignored in the decision-making process.

Figure 10: The CENELEC EN 50600 Data Centre Infrastructure Standards

Infrastructure Impact of Administrative Class 1 Class 2 Class 3 Class 4

Component short failure risk

Power supply and No Concurrently

Distribution No Resilience Maintainable Fault Tolerant
HIGH MEDIUM UPS (Single
N+1
2(N+1)
(Active/Passive)
EN 50600-2-2 Path)

Environmental Control No
No Concurrently Fault Tolerant
LOW MEDIUM Resilience Maintainable
EN 50600-2-3 ECS N+1 2(N+1)
(Single (Active/Passive)
Path)

Telecomms Cabling Concurrently

No Resilience No Concurrently Maintenance
EN 50600-2-4
MEDIUM HIGH (Point-to-Point Resilience Maintenance (Diverse
Cabling) (Fixed Cabling) (Multi-Path) Routed)
Lowest Risk

Highest Risk
Highest Cost
Higher Cost
Higher Cost
Lowest Cost

20
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

STRUCTURED OR NOT? CABLING SPECIFICATIONS

The broad advantages of structured cabling over point to point The choice of the right networking equipment is a matter of piecing
cabling (such as that used on top of rack configurations in a data together different elements of standards, technical specifications
centre) are seen as: and physical properties. This first set of specifications is used to
illustrate this process for copper twisted pair cabling:
improved energy efficiency
increased switch port utilisation The different IEEE standards for copper twisted pair cabling are shown
quicker system adaption at the top of each column with the cabling category and the effective
greater scalability length that the transmission will carry across although it should be
easier network expansion noted that the rules of link length can be overcome with a growing
less maintenance and administration range of connector choice for both copper and fibre that can extend
less expensive switches the reach of the cable with minimal compromise to the signal.
overall lower cost of ownership.
The baud rate and channel bandwidth indicate the speed and
Disadvantages cited include the fact that more cables and racks capacity of traffic that will be able to pass through the network
are needed leading to higher cross-connect copper cabling across the particular cabling specification.
costs, lower suitability for high density blade servers and length
restrictions on cross-connect copper cabling. The specifications are increasingly important as the development
of higher speeds and bandwidths is required to accommodate the
The structured cabling system can make a network easier to growth of traffic within data centres. Data transmission rates in data
manage, and make it easier also to identify and deal with specific centres have increased for copper from 1G to 10G to 40G plus 1G
points of failure. It divides up the network into manageable blocks then 5G to 25G, and for fibre, from 10G to 40G plus 25G to 50G
and works on a defined hierarchy of management. The data to 100G. This rise in data rates means an increase in bandwidth
centre is divided into distribution cabling sub-systems main, requirement. The minimization and reduction of cross-talk and
intermediate and local. The main distribution cabling sub-system interference to transmission is also a key measure of performance.
will house the main cross-connects and the core networking While this may partly be down to installation, this forms part of the
equipment. Intermediate sub-systems will house the cross- core product specification. The decibel measures shown represent
connect to the equipment areas. Industry consensus suggests the expected minimum operating environment.
structured cabling system makes additions and modifications
easier to achieve and allows for flexibility in connections. Figure 11: IEEE Technical Specifications for Multi-Gigabit Protocols
The deployment of data centre modules is increasingly common 10GBASE-T 40GBASE-T 25GBASE-T
as a means of building data centres based on the principle that Cabling 100m Cat 6A 30m Cat8 30m Cat8
it is more cost efficient to construct and fit out a data centre
Baud Rate 800Mb 3200Mb 2000Mb
as demand grows. The nature of structured cabling and the
use of standards and specifications agreed between suppliers RX-ENOB (bits) 9.5-10 7.8 6.5-7.5
and industry associations has to some extent already created
IL 46.9 29.4 22.6
a modularity within network and cabling systems, which means
they can be connected and upgraded as the requirement for the Channel Bandwidth 500MHz 2000MHz 1000-1250MHz tbc
network changes. Cabling systems now take the form of modular Echo Cancellation 55dB 55dB 55dB
plug-in/snap-in systems where the cabling may have been pre-
Next Cancellation 40dB 34-40dB 28-40dB
terminated and pre-tested before leaving the factory.
Fext Cancellation 25dB 22-25dB 19-25dB

Relative SNR Margins Reference Phy +2.7 to 0.2 +8.7 to +4dB

Relative Power Reference Phy 2.5x to 3x 10Base-T 1.3-2x to 3x 10Base-T

21
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

The same principles of balancing transmission capacity and speed
requirements against the distance at which the signal will maintain
integrity are applied also to the specification of fibre cabling
(shown on the grid below). Loss of signal between the near and
far end of the line applies also to fibre cabling but depends on the
transmission speed, the distance and the number of connections in
the link. Modal bandwidth indicates the capacity of multimode fibre
cabling to transmit a signal at a maximum rate over a given distance.
The gradation from OM3 through OM4 has occurred to increase
the modal bandwidth to provide extended distance support for
bandwidth limited applications. The proposed move to OM5 is
designed to increase the rate of transmission and provide higher
data throughput.

Figure 12: Supported Distances for Applications on Singlemode and Multimode Optical Fibre Cabling
APPLICATIONS AND PERFORMANCE
Nexans LANmark-OF OM4 MTP fibre systems are warranted to support following applications.
Guaranteed distances vary depending on application and number of connections.

ETHERNET DISTANCES-DUPLEX TRANSMISSION

# MTP-L-C 2 4 6 8

1Gbase-SX 930m 910m 890m 830m

10Gbase-SR 520m 510m 500m 470m

25Gbase-SR 105m 95m 85m 65m

40G BIDI 145m 140m 135m 125m

ETHERNET DISTANCES-PARALLEL TRANSMISSION

# MTP CONNECTIONS 2 4 6 8

40Gbase-SR4 170m 160m 150m 150m

40Gbase-SR4 extended distances* 400m 360m 330m 300m

100Gbase-SR4 120m 110m 110m 110m

* Nexans approved transceivers

ETHERNET DISTANCES-PARALLEL TRANSMISSION

# MTP-LC MODULES 2 4 6 8

1GFC (PI-4 100-M5F-SN-I) 1150m 1120m 1050m 950m

2GFC (PI-4 200-M5F-SN-I) 680m 660m 620m 550m

4GFC (PI-5 400-M5F-SN-I) 470m 450m 430m 380m

8GFC (PI-5 800-M5F-SN-I) 220m 215m 200m 160m

16GFC (PI-6 1600-M5F-SN-I) 150m 145m 130m 100m

32GFC (PI-4 3200-M5F-SN-I) 100m 95m 90m 75m

10GFC (10GFC 1200-M5F-SN-I) 520m 510m 500m 470m

Applications are only warranted if LANmark-OF OM4 patchcords are used.

22
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

COPPER, FIBRE OR BOTH?
The key cabling profiles of copper and fibre shown previously
can be deployed to meet the different networking needs of the
different spaces in a data centre building.
The first requirement is to understand the bandwidth
requirements of each zone. This will depend upon the density
of the servers and the networking equipment within the data
centre, and in other areas of the building. This will determine the
category of copper or fibre cabling required. It indicates also the
maximum length of the cable that will transmit at the specified
bandwidth. Wireless links will be an option in the office or control
space where the amount of traffic is far lower than in the data
centre or server room. The development of wireless transmission
within and from the data centre is still very much at an early stage.
Wireless technologies potentially offer the capability of short
bursts of added network capacity at peak periods, and they can
be switched on and off as required. (11) (12) This indicates the
importance of a strategy that differentiates between the different
requirements of different zones within a data centre building.
The choice between copper and fibre cabling is one of the most
debated in the data centre industry; the decision on which is more
suitable will depend on a number of considerations. However, the
different technical characteristics and physical tolerances of each
type, particularly in terms of the distance at which the link can
maintain integrity of transmission tends to define the decision.
Fibre offers transmission over longer distances within and outside
the data centre particularly in single- rather than multi-mode. It
is seen to offer greater bandwidth and data rates, immunity to
electromagnetic interference, being more space efficient, lighter
and generally quicker to install.
Copper has been preferred within the data centre as the server
link between the equipment and the access switching. Copper is
used to connect the switch to the server and the network rows to
the server rows as part of Middle of Row and End of Row network
row connections. In this role it is used to link up to 30 metres
across server rows.

Figure 13: How Data Centre Building Spaces Differ in their Network Cabling Requirements
COMPUTER ROOM SPACE
SMART OFFICE SPACE OTHER SPACES
CHOICES
SWITCH TO SWITCH SWITCH TO SERVER

Low High Very High Low High Very High Low High Low High
Bandwidth
<10Gbs-1 <25Gbs-1 <40Gbs-1 <100Gbs-1 <10Gbs-1 <25Gbs-1 <40Gbs-1 <100Gbs-1 <1Gbs-1 <10Gbs-1 <10Gbs-1 <1Gbs-1 <10Gbs-1

LANmark LANmark LANmark LANmark LANmark LANmark LANmark LANmark LANmark LANmark LANmark
Copper
6A 7A 8 6A 7A 8 6 6A 7A 6 6A

Distance
100m 30m 100m 30m 100m 100m
(Copper)

Fibre OM3 OM3 OM3 OM3

Distance
350 85 300 85 350 85 300 85 900 340 340 900 340
OM3

OM4 OM4 OM4 OM4

Distance
550 120 400 120 550 120 400 120 930 520 520 930 520
OM4

Other issues
Roll-Out
Low High Very High Low High Very High Low High Low High
Speed

Roll-Out On-Site Plug and On-Site Plug and On-Site Plug and
Pre-Term Pre-Term Pre-Term Pre-Term
Technique Term Play Term Play Term Play

LANmark is a registered trademark of Nexans.

23
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

It is important in this process to achieve the right balance in

Figure 14: Reach and Speed Specifications of
provisioning and to design switch-to-switch cabling layout and Copper and Fibre Network Cabling
bandwidth according to performance needs whether the access
is Top of Rack, Middle of Row or End of Row. While uplinks between
switches may require speeds of 100G the switch to server downlinks
Metropolitan
will require lower capacities of 40G or below. This can be more cost- 10Km SINGLEMODE FIBRE
effectively provided using structured copper cabling.

Copper can also be considered for End of Row uplinks where the
Building-based
distance is less than 30 metres. Using Category 8 as an upgrade 1Km
MUL
TIMO

Reach
path means switches can be upgraded in steps from 1G to 40G DE F
IBRE
without re-cabling and the number of switch ports can be reduced
Room-based
100m COPPER TWISTED PAIR
by a factor of 10. At 40G this can lower energy use and costs by as
much as 50% and quadruple link speed. The cost of EoR access
Row-based
using 40GBase-T (over copper cabling) is around one-third of that 30m
of EoR with 40GBase-SR (over fibre cabling), and lower again than COPPER Rack-based
TWINAX
ToR with 40GBase-T and ToR with DAC. Transceivers account 5m

for 60% to 70% of the costs of the latter three options, and the 10M 100M 1G 10G 100G
equipment costs are similar in all four options. The advantages in
Speed
this usage situation is that a Base-T Ethernet approach reduces the Source: Alan Flatman
cost of energy; transceivers are less expensive, the components
are high volume and relatively easy to manufacture. It is easy to
install using field termination and is available in pre-terminated
Figure 15: Copper Network Cabling Usage in Different Local
Distribution Connections
variances, backward compatibility and plug and play and the reach
will be adequate to a maximum of 100 metres for bandwidth up to
Overhead Row-to-Row/ Overhead In-Row Trunks
10Gb/s or 30 metres for bandwidth of more than 10Gb/s. Pod-to-Pod Trunks Generally Copper
Either Copper or Fibre depending Structured Cabling
In the modern data centre, the need to prepare and support on distance Structured cabling
several generations of active equipment mean the specification
and standard of the cabling and network equipment needs to plan
for future networking requirements. Some provision in the form of
spare connection points, patch panels, links and ports or through
devices that can scale quickly to a future requirement has to be
built into the topology.

In the data centre there is likely also to be a premium on space

and the need to migrate to higher speeds or to add ports without
disrupting operations. Therefore density, accessibility and
scalability are of high importance. To meet these requirements,
several solutions can be helpful, such as angled patch panels to
ease access to patch cords, or high density fibre panels which
Under-floor
are able to increase the numbers of connections in a U of rack In-Row
space or bend-insensitive fibre patch cords that have smaller Underfloor Row-to-Row/ Trunks
bend radii and can accept inclusion in smaller spaces. Pod-to-Pod Trunks Generally
Either Copper or Fibre Switch Port Harness Copper
depending on distance Generally Copper Structured
Structured cabling equipment cords Cabling

24
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

STEP 4 OF 5

INSTALLATION
BEST NETWORK CABLING AND
EQUIPMENT INSTALLATION PRACTICES

There are a number of key contributions to best practice related

to the installation of cabling in the data centre:

The pre-existing state of management of cabling and network equipment,

and the ease of the change or repair identification. Most data centres have

1 allotted periods of planned downtime and one of the major causes of

unplanned downtime is unexpected repair, maintenance or upgrade work.
This means best practice in network documentation and organisation is critical
to data centre availability.

Equipment manufacturers can contribute to a smoother and more accurate

installation process. This can be done in two ways. First, equipment should
2 be designed to ensure it can be installed easily and correctly. Second, the
equipment should be supplied to the data centre in the order in which it will
be installed and close to the time when that will take place.

The use of sub-contractors who are fully accredited and trained to the necessary
installation standards (e.g. EN 50174) and who are also conversant with the
specific needs and conditions of the individual data centre. The sub-contractors
3 should be in full contact with the manufacturer of the equipment, to ensure
the installation process is as seamless as possible. It is critical the process is
conducted according to a Statement of Work which is agreed and signed off
by all parties.

Proper expectations established with clients about the length of time required
for the installation process including testing, and proper preparation of the site
4 beforehand. While in many data centres planned downtime costs availability
and therefore money, the need to install fast may put undue pressure on the
people and lead to unwanted outcomes.

25
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING
The key requirement of cabling installation is that it should be
quick and easy, fully organised and accurate in its outcomes. The
process needs also to conform to the conditions of equipment
supplier warranties. A well-managed installation process will be
conducted by specialist contractors who are aware of the physical
limitations of cabling such as bend stress and who will avoid over-
bundling or crushing cables as well as being fully updated on local
space, access and air flow requirements. The installation process
should also account for future cabling installations and also avoid
installation of cables in hot areas or close to power cabling.
One of the key aims of cabling installation is to standardise the
process for cables, switches, panels, cabling management as
well as for labelling, colours and documentation. This process
should also establish the pathways and conduits through the data
centre, the design and installation of connections between the
sequential distribution systems within the overall system (main to
intermediate, intermediate to local). Largely, the standardisation
should be established at the design stage so that the work on-
site is a matter only of execution rather than making on-the-job
decisions as to how to progress. Like any other work conducted
in the data centre, the process needs to conform also to
occupational health and safety requirements.
Recent innovations by network equipment designers and
manufacturers have created products that can assist in the
installation process. Ideally, solutions need to be designed so that
Figure 16: A Typical Ultra-High Density Fibre Panel
26
www.dcd-intelligence.com
they can be installed quickly and easily through pre-installation
design rather than extensive on-site work. These solutions include:
Cassette format with neatly packaged fan-outs to avoid
tangled connections
Pre-terminated cable assemblies
Cables delivered at individually specified lengths
Housing or locking as security is important even in a location
with very limited or no public access such as a data centre.
A number of particular innovations relate to the installation
of fibre cabling:
A specialist containment system will protect fibre against
damage
Pre-installed MTP-LC modules. These can save as much
as 70% of time required on site compared to traditional
termination methods
Patch panels that allow for modules and enable fingertip
access to patch cords in separate trays
Colour matching of fibre pair-flips.
The manufacturer will also be able to recommend on the stocking
of spare cabling products. As installation progresses, remove old or
unwanted cabling and equipment from the data centre to reduce
the fire load and avoid the possibility that it might cause confusion
in the future.
 DATA CENTRE
NETWORK CABLING

THE IMPORTANCE OF TESTING There is no set time that

a process of network
THERE IS NO
It is important also that cables being installed are tested in installation should take. For
SET TIME THAT
order to ensure their quality and performance. Such testing is each individual project there
usually conducted by an authorised specialist and in accordance
A PROCESS are different requirements,
with guidelines and field test procedures as issued by the
OF NETWORK different local conditions
manufacturer. To achieve best practice in testing: INSTALLATION and considerations. Yet
SHOULD TAKE. FOR the process can be made
Only use test equipment that has valid annual calibration EACH INDIVIDUAL easier and quicker through
certification PROJECT THERE completing as much of
Make sure that the correct test leads are used, again ARE DIFFERENT it as possible before the
following the manufacturer recommendation REQUIREMENTS, equipment reaches the data
Calibrate the tester as required during the testing phase and DIFFERENT LOCAL centre. And by using the
ensure that the firmware and software database are updated CONDITIONS AND installation process to ensure
Verify the correct testing method and ensure that the tester that sufficient allowance is
CONSIDERATIONS.
can accurately diagnose pass from fail made for future installation
Use Light Source Power Meter (LPSM) and as required the requirements and through
Optical Time Domain Reflectometer (OTDR) for fibre links. the proper management of the project once completed, this will
mean that installing network equipment should get easier. But as
The correct testing process can then be used to apply for a with all other aspects of cabling in the data centre, the direct flow
system warranty from the manufacturer. of decision making from overall strategy through the technical
expertise of manufacturers and consultants will ensure the
It is critical that all work done is fully documented as part of the installation process meets required outcomes.
system of data centre inventory management. This will record all
additions or changes made to cabling and networks in the data
centre. While the documentation needs to be supplied by the
contractor who has installed the cabling, it needs to be owned
and updated by data centre staff.

27
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

STEP 5 OF 5

MANAGEMENT
BEST DATA CENTRE NETWORK
CABLING MANAGEMENT PRACTICES

A network system that is well managed and documented is easier to future-proof

and modify.

Among the key best practice components in data centre cabling management are
documentation and inventory control that can be updated with changes to the network
or to related equipment. Ideally, these components will allow relevant information to be
shared as necessary with staff or contractors. It will be important also to establish protocols
as to dealing with the network system equipment and its equipment once standardisation
is established as well as to how the information collected should be used, shared and
acted on. These should cover installation, repair, maintenance, upgrades and recovery.

Figure 17: Colour Coded Products for Easy Identification of Services

28
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

The process of network equipment Management programmes need to be

management can be facilitated by the MANAGEMENT established for the cabling and networking
colour-coding and through the labelling PROGRAMMES NEED equipment to ensure equipment meets
of cables, ports on patch panels, and TO BE ESTABLISHED the network requirements for which it was
connections. A system for patch cables FOR THE CABLING designed. These programs should also identify
might, for example use different colours to AND NETWORKING threats to achieving this and strategies for
distinguish between the services that the eliminating those threats. Where breakdowns or
EQUIPMENT TO
cabling supports. Cabling, even shorter failure occur, the reasons need to be identified.
ENSURE EQUIPMENT
links, should be labelled at both ends. Inspection and testing can be scheduled and
MEETS THE NETWORK undertaken, as required an approach which
EMAC (Environmental Monitoring and REQUIREMENTS may help budgeting, equipment longevity, and
Access Control) systems can be configured FOR WHICH IT WAS the reduction of failure rates. However, any
to activate alarms when tolerance DESIGNED. THESE management system does need to be flexible
thresholds are reached. The monitoring and PROGRAMS SHOULD and adapt to what works most effectively in the
access control system can identify threats ALSO IDENTIFY THREATS context of the equipment and its performance.
to both the data centre and to the network TO ACHIEVING THIS
within it. These factors include: AND STRATEGIES FOR As part of the best practice management
ELIMINATING THOSE of cabling and network systems, security
power management, THREATS. systems need to be in place to restrict physical
temperature access to the site and areas of it. This is of
humidity and security sensing; particular importance to shared data centres
management of important cabling links where there may be a number of different organisations or units
connections and devices; within an organisation requiring access to their own aisles or
change control and work orders for both cabling and non- racks. A number of site security issues can be overcome with
cabling tasks as part of maintenance best practice. comprehensive online access to live views of the data centre,
and with monitoring systems that identify threats and activate
In relation to cabling, the system can also protect circuits in alarms where necessary. Cabling and network equipment will be
the patch zone from accidental disconnection and deliberate part of the overall data centre or zone security, locks and housing
interference with security devices. systems can be used to reduce network threats.

Figure 18: Examples of Secure Lock to be Used on Copper or Fibre Patch Cords

29
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

WHAT TO DO NEXT

This white paper has been written on behalf of Nexans.

In the field of LAN cabling systems, Nexans offers a complete range of products and
value-added services providing improved reliability and reduced cost of ownership for
data centres, offices and campus networks.

Nexans LANmark cabling systems allow you to optimally balance copper and fibre, in
line with the performance you need, for any number of ports and any bandwidth. High-
performance LANmark connectivity with minimised insertion loss allows for greater lengths
and more connections without sacrificing quality or reliability.

Our connectivity solutions exceed the requirements of all relevant standards, ensuring your
data centre performs better and more reliably. Whats more, Nexans expertise and far-reaching
involvement in the development of new protocols and standards for data centres and cabling
means we find the right solution every time.

Our design and manufacturing innovation allows us to offer solutions that meet and exceed
the requirements of todays cutting-edge data centres in terms of cost, space, installation and
energy efficiency and that help future-proof the data centre in terms of future connectivity
requirements with a minimum of disruption.

30
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

Nexans data centre solutions include:

Flexible, cost-effective bandwidth architectures
Optimized space usage
Performance protection and optimization
Faster modular deployment
Design through to operation support services

These are provided to a global customer base through

an extensive network of regional offices and Key Account
Management team.

Please contact/visit www.nexans.com/LANsystems to find out

more about Nexans own commitment to best practice in data
centre connectivity.

Figure 19: Nexans Delivery Capabilities

How Engage can help

PLAN
Keeping you up-to-date on the latest innovations and Reducing Total
Cost of Ownership
helping plan for future changes.

DEFINE
Sharing our expertise in defining state-of-the-art
specifications and solutions.
Enabling Meeting higher
DESIGN sustainability bandwidth needs
Supporting you in designing robust, flexible
and scalable systems.

DEPLOY
Blueprinting infrastructure for cost-effective and accurate
installation, assisting on-site and ensuring warranties.

USE
Offering training, monitoring and fast repair, Anticipating Improving
technology change network reliability
replacement and third-level support.

31
www.dcd-intelligence.com
 DATA CENTRE
NETWORK CABLING

BIBLIOGRAPHY

(1) Cisco Global Cloud Index (GCI): Forecast and Methodology 2014 2019;
October 2015.
(2) DCD Intelligence Global Census 2014-2015; findings previously unpublished.
(3) Global Data Center Market Overview and Forecasts 2014-2020; DCD Intelligence;
April 2015.
(4) Uptime Institute: Operations and Management Annual: October 2013.
(5) Data Center Outages: The Ponemon Institute and Emerson Network Power:
2013 Study on Data Center Outages.
(6) Data Center Research Group; December 2011.
(7) DCD Intelligence Global Census 2014-2015; findings previously unpublished.
(8) DCD Intelligence Global Census 2014-2015; findings previously unpublished.
(9) IEEE P802,3bq 40GBASE-T Task Force, presented Beijing China; March 2014.
(10) Communications Cable & Connectivity Association; quoted Todd Harpel, RCDD,
Marketing Committee Chair; BISCI.
(11) Project Genome: Wireless Sensor Network for Data Center Cooling;
The Architecture Journal: January 2009.
(12) Relieving hotspots in data center networks with wireless neighborways;
IEEE; December 2014.

32
www.dcd-intelligence.com
DatacenterDynamics (DCD), part of the DCD Group, is an
international B2B media and publishing company. It offers a
comprehensive range of products and services to assist senior
professionals in the worlds most ICT-dependent organizations in FOR MORE INFORMATION:
their risk-based infrastructure and capacity decisions.

DCDs portfolio of live events, online and print publishing, business

intelligence and professional development brands are centered +44 (0)20 7426 4813
around the complexities of technological convergence. Accessible
from over 42 countries, DCDs unique knowledge and networking info@dcd-intelligence.com
platform is trusted by over 40,000 ICT, Engineering and Technology
professionals worldwide. www.dcd-intelligence.com