News

The choice of cabinets can help simplify overall data centre management. This is the case if data centre managers opt for pre-configured cabinet solutions. These cabinets are preloaded with components, such as fibre or copper connectivity, power distribution units (PDUs), cable management or other accessories. They are pre-assembled, packaged and arrive on site only requiring final connections and installation of active equipment. The ‘purpose’ of each cabinet  – e.g. a server, storage area network or a networking cabinet – can thereby be pre-designed upfront but cabinet design can also be based on capacity, e.g. 5kW or 10kW.

And because each preconfigured cabinet is identified by one unique customer-specific part number and price, the ordering process of future identical cabinets is much more simplified.

In addition, preconfigured cabinets achieve 30 per cent savings on time and labour compared to traditional cabinets where each component has to be unpacked and installed.

The choice of cabling will determine whether the data centre can actually facilitate the ever increasing demand for higher Ethernet speeds of today’s dynamic businesses. At the data centre edge, switch-to-server connections now require 40 Gb/s and even higher speeds and this has resulted in the devleopment of 25GBASE-T and 40GBASE-T Ethernet applications for this part of the data centre. To satisfy the 2 GHz bandwidth requirements to support 25 Gb/s and 40 Gb/s throughput, a new category of cabiling -Category 8 – was developed. Today, class I, class II, and category 8 cabling are characterised to 2 GHz supporting 30 metre cabling channels that contain no more than 2 connectors.  These channels and the emerging 25G/40GBASE-T applications that they support are specifically targeted for deployment at the data centre ‘edge’ where server to switch connections are made. 

In the data centre backbone, speeds are now surpassing the 100Gb/s mark and 200 and 400 Gigabit Ethernet speeds will soon be supporting bandwidth hungry applications. Here it is important to ensure that the installed data centre cabling technology can support migration to these future applications.

With 200 and 400 Gb/s applications transmitting over fibre counts that are divisible by 2 or 8, deployment of new 8-fibre/MTP solutions – instead of the commonly installed 12-fibre MTP connectivity – are highly recommended. Using 12-fibre MTP connectivity to support current and future 8-fibre applications would mean that only 8 of 12 fibres will be in use, with four fibres or 33 per cent going unused. Using 8-fibre MTP backbone cabling and 8-fibre MTP jumpers instead, achieves 100 per cent fibre utilisation.

Today, 8-fibre MTP solutions provide the most efficient, cost-effective and highest performing option for current 8-fibre 40 and 100 Gigabit applications such as 40GBASE-SR4 and 100GBASE-SR4, and they also provide an easy migration path to next generation 200 Gb/s and 400 Gb/s applications.

For data centre facilities that already have 12-fibre MTP solutions installed and are looking to support current and future 8-fibre applications, conversion cords or modules can be deployed that transition two 12-fibre MTPs to three 8-fibre MTPs to enable 100 per cent fibre utilisation. This however is a more complex way to achieve 100 per cent fibre utilisation compared to using 8-fibre MTP solutions.  Conversion cords mean that 3 ports will have to be taken offline in the event of a damaged cord, while conversion modules introduce additional insertion loss into the channel which might impact overall performance.

The rising demand for computing power also requires data centre managers to expand their facilities quickly and cost effectively and the way in which data centres are being designed can effectively contribute to speeding up this process.

Traditionally, data centre growth has been addressed on an as-needed basis and cabinets are typically piecemealed together based on an unplanned, ad hoc design. Compare this to a more modern, modular data centre design approach that uses Pods. These pods consist of groups of cabinets that are typically based on capacity, function or application and are typically connected to the network’s core switches via aggregation layer switches that may reside within the pod or in a separate distribution area that serves multiple pods. Pods are often comprised of two symmetrical rows of cabinets configured in a hot/cold aisle layout and typically deployed using hot or cold aisle containment solutions. For example, a pod could contain two rows of 6 cabinets that are placed back to back in a hot-aisle containment configuration or placed front to front in a cold-aisle containment configuration. Pods can be easily repeated as the data centre grows, allowing customers to scale the data centre in a more predictable manner. The initial pod that is designed and deployed on day-one serves as a template for incremental build-outs of additional pods. Because a modular pod-based design is repeatable with consistent cabinets and components replicated from pod to pod, it provides a predictable set of costs and resources required for planned data centre expansions. The initial pod’s deployment process, as well as network, compute, storage, power, cooling and space requirements, can be repeated for future pods. Data centre managers hence know what to expect, reducing the time and risk associated with traditional ad hoc expansions and upgrades.

Design considerations at the physical layer play a crucial role in a data centre’s ability to keep up with the growing demands and speedy changes of the business it serves. Careful choice of design approach and components will ensure that a data centre facility will remain flexible, scalable and manageable in line with what the business demands from it.

Nicolas Roussel
Technical Manager Central Europe
Siemon