Made in European Union

Correct cabling – but how?

What people in charge should care about planning a DC project


Like all other things which should be good in the end the planning is a basic element in data centers. As well as infrastructure and construction requests cabling is also important. Suitable to operation and sustainability planning also has to take into account breakdowns and service. So the most important questions for the planer are: What do we need for what and where should it go? An planning example.
It should be taken into account which components are currently in use and which must be provided in the following years. When selecting the components not only the initial installation is important but also the consideration of the ongoing operation as well as the generation change of the active systems after about three to four years. For this case the cabling must be simple and practically expandable. Here space is required in the cabinets for future-proof expansion because components cannot be removed simply and new ones inserted but rather embedded in operation in parallel. Systems are ideally economically time-saving, can plug-and-play and are ready for use and serviced in a few simple steps.

Planning

One of the first steps from the point of view of the cabling is therefore the decision whether the data center is to be implemented in single mode in the fiber optic area or in a multimode structure. The singlemode fibers are significantly cheaper than the multimode fibers. The connection components for singlemode fibers are a bit more expensive. Higher bandwidths can be channeled via singlemode fibers, but the active components are more expensive and the trend is falling. Therefore the decision whether singlemode or multimode should not only be considered technically but also commercially.

However the multimode fiber is still not to be neglected: In recent years OM5 has become a keyword, even if not many projects in the OM5 area have been implemented so far. Current OM4 cabling shows similar values to their OM5 counterparts and are usually cheaper. The key difference: New applications on OM5 are becoming the norm. Because: So far multimode fibers have only been operated with one optical window at 850 or 1,300 nm. In order to secure the increasing bandwidth the new fiber type OM5 with extended transmission properties was developed. This has been extended by three further optical transmission windows (880, 910 and 940 nm) in order to use SWDM (Short Wave Division Multiplexing). SWDM is a technology that enables four transmission channels in one fiber pair. In touch with the MPO connector (new types of this connector are currently being developed), it will be possible in the future to achieve transmissions of 800 Gbit / s.

In multimode and singlemode the MPO connector, also known as the MTP connector, and the LC Duplex are common standards. Both plugs are offered as singlemode and multimode. However the LSH or E2000 is very popular because it is easy to use and has good mechanical properties. Both types are common as PC or APC connectors, so they vary in the type of contact (physical contact). The face of the APC connector (Angled Physical Contact) is inclined by eight degrees so that the return loss is kept low and thus more attenuation buffer is available. In singlemode the MPO is always an APC connector with the LC available as either a PC or an APC.

Operation

The planning of current operations also includes handling in the event of a Breakdown. Breakdowns cannot be excluded even with the best network maintenance and must therefore be diagnosed and rectified quickly. Service and downtime can be very expensive. Even with redundant data centers the following applies: If the active side is blocked, no further service can be carried out there. In the worst case if an error occurs there, the result is failure. Time is a critical factor so error diagnostics and fault repair must be dealt with very quickly.

Faults on the hardware side are quite easy to identify and technicians can read out errors directly on the machine. However if the fault is on the cabling side, the line must always be traced locally. Has the patch been correctly patched, have all contacts been connected correctly and are surfaces clean? The last point in particular is often underestimated. Cabling structures are sensitive, high-quality products that allow and require a certain standard of cleanliness. Components such as LED systems in the cables facilitate and accelerate the search for connected cable ends in the event of a fault. In this way you reduce lost time and costs in the event of a fault.
Another element on the cabling side of smooth operation is the packing density. If the packing densities are high a flexible modular design is recommended. In this way individual modules in the structure can be exchanged. Of course, thinner cables also counteract the packing density. This must be taken into account when planning the cable management. During the initial installation high-fiber cables can be used in the hot aisles in the doubled floor. Post and patch cabling can then be done above or inside the cabinets.

While a lot of cables used to be required due to point-to-point connections and were laid in doubled floors these floors are no longer part of the current standard. The difficulty was that if there were too many cables lying on top of each other the lower ones could no longer or only poorly be replaced. At the same time they encumber the flow of air. Today cable lines are based on the cold and hot aisles: warm and cold air is used in such way so the space can be used for basic cabling with thin and high-fiber cables in the hot aisles. Another advantage of such slim lines is that they do not obstruct the flow of air.
Cable management within the data center also needs to be taken into account at an early stage: Are there enough redundant paths available here that are also attached in such a way that maintenance is ensured without switching off the devices? Are the excess lengths taken into account in order to avoid unnecessarily coiled cables? The cable management is also not the same in all cabinets. Some cabinets are designed for servers and others for networks. The cabling structure is important here: not every cabinet requires basic cabling. For example only every third or fifth cabinet in a block can be approached. This results short patch cables from cabinet to cabinet and by the way lower costs.

Sustainability

It goes without saying that a look into the future is also recommended for cabling. This does not just mean a certain flexibility and adaptability to changing standards. The demands on data centers are increasing due to greater compatibility with existing networks and transmission speeds. New Cat.8.1 cabling systems are emerging even if active end devices that require such data volumes are not yet common enough. But: A change in a few years could be costly. It therefore makes sense to be prepared for future cabling requirements by means of appropriate module technology and selection of cable types and to be able to avoid re-assembly afterwards.
If a conversion is necessary it should ideally be easy to implement. In data centers that use a redox system to prevent fire, for example, service personnel must be medically examined as a result of the oxygen reduction and must not stay too long in the data center. Here it makes sense to use a very flexible structure including components the basic structure of which can be changed without great technical effort.
Ultimately future sustainability is also significantly influenced by supposedly minor things such as service. An often underestimated aspect in this context is the cleaning of the cable structures. Too often cables are only patched. Glass fiber in particular is very sensitive to dirt. With increasing data rates the applications become more and more sensitive.
Contamination can then lead to connection failures. This was shown in the SAN area with the change from 4 Gbit / s to 8 and 16 Gbit / s. Long term the following questions will help: How were the cables laid? Is the necessary cleaning and service equipment available? Are the employees trained to such an extent that they also clean and check the existing system before commissioning? In the end no one in charge is helped if the data center is perfectly planned and the IT infrastructure is ultra-modern but errors in the cable management disrupt operations.

Writer: Stephan Elfe

 

Back