The next wave of cabling
Networking cabling infrastructures need to evolve in this age of digital transformation. Network Middle East examines what the key considerations are to ensure least disruption to business before upgrading data centre networks
Networking is changing at a fast pace in the data centre. The network architectures that make cloud possible are fundamentally different from the highly over-subscribed, hierarchical and costly legacy solutions of the past. In 2020, advanced technologies like 5G are expected to be applied in the data centre, offering data centre providers greater opportunity to grow and enhance their business. While the benefits of these technologies may take at least a few years to come to fruition, data centre companies that incorporate them into their business strategy now will be the ones best positioned to reap the benefits down the road.
The deployment of machine learning, deep learning and other artificial intelligence technologies is now mainstream, and these technologies power many of the cloud services we use every day, says Alastair Waite, data centre market development, Commscope. “Plan for change from day one, build with maximum flexibility and choose a good installation partner to support you, ” he adds.
Networking cabling infrastructures in the DX age
Increased adoption of high-performance servers and applications requiring higher bandwidth is driving adoption of 10 and 25 Gigabit Ethernet switching in combination with 40 and 100 Gigabit Ethernet. As a result, it has become increasingly necessary to include network fabrics, software-defined networking, hyper-convergence, and software-defined storage technologies in the data centre. From a cabling standpoint, all this added complexity and modernisation of the data centre promotes the need for simplicity, reliability and high-density.
To plan an upgrade or a new technology leap within the customer network, the technology of choice should have a proven stability record, be suitable for the new applications that keep coming out, and should be future-ready. Also, it should be clear how the technology can be implemented or upgraded. In general, customers in the Middle East are aware of the importance of digital transformation and organisations are increasingly taking steps to be prepared.
Customers need a clear vision of what the final results should be and how to structure the upgrade process, to ensure all steps are under control and disruptions are minimised, explains Arafat Yousef, managing director MEA, Nexans Cabling Solutions.
“It makes sense to choose solutions that can be upgraded on a modular basis, allowing the customer to add or change features in line with requirements. Things become even better when a solution allows changes to be made at components level, while the key frameof-work remains fixed. This means better preparation for future expansions with minimal disruptions and acceptable cost,” adds Yousef. “We always find big crowds attending seminars and roadshows by all vendors, across various areas of interest. Here, people look forward to getting the latest updates about standards and industry trends, to plan upgrades to their current networks and the direction of new projects accordingly.”
Internal data centre traffic is expected to grow by 80% over the next three years. Because of this, there is a real risk of networks becoming bandwidth bottlenecks. Reports indicate that 65% of system outages are related to cabling and that patching mistakes are the reason for 28% of downtime in data centres. However, when planning the installation or upgrade of network cabling, the actual cost of cabling typically only accounts for just 4-5% of the total expense of the data centre.
As a standard practice, organisations must move away from traditional lowdensity cabling to high-density modular structured cable solutions, says Nabil Khalil, executive vice president of R&M META. “By doing so, they can implement physical network infrastructure in a far more manageable and flexible manner. Furthermore, these systems enable data centres to easily migrate to 25, 100 and 200 Gb/s networks and solve some of the most critical network challenges.”
Fibre: To be or not to be
With the new Category 8 standards capable of supporting up to 40 GBE, many are left undecided if there is still scope for copper in the data centre or whether fibre is the only way forward.
Fibre optic cabling offers several advantages over copper such as higher throughput, space savings, better security and future-proofing while copper cabling can only be deployed when it can reliably meet the requirements of a specific application. However, one reason why organisations still favour copper is that they tend to be a cheaper alternative to fibre.
Waite explains that there is still scope and demand for structured copper cabling in the data centre. “Customers still rely on copper to provide networks for out-of-band management of servers and switches, wireless access points, security access systems, and the many IOT sensors that are located within the cabinet, cage and all across the data hall to ensure the smooth and efficient running of the white space. New advances in POE technology have also ensured that copper remains a relevant transport media all over the data centre campus.”
Non-hyperscale data centres still use a great deal of copper cabling to connect servers. Economically, this makes perfect sense. Copper Twisted pair runs at 1Gbps or 10Gbps today. 25Gbase-t and 40Gbase-t are defined as the next steps in Ethernet standards but are not yet deployed to the market, as the bandwidth demand in non-hyperscale data centres can still be satisfied with 10G per port.
When it comes to future-readiness, fibre is currently hyperscale data centres’ first choice, adds Yousef of Nexans. “However, when the need for higher speeds arrives, new cabling will clearly be required: Cat6a will no longer suffice.”
Nexans recommends using Cat7a cabling as soon as twisted pair is capable of running beyond 10G. Cat7a supports 25GBASE-T. Cat8 cabling is required for 40GBASE-T, but the economics of system design have already made this speed obsolete. “We expect only 25Gbase-t to hit the market in the future. Cat8 would be usable, but Cat7a is already sufficient and more affordable.
R&M VP Khalil explains that copper and fibre both hold their ground depending on specific use cases. “Copper is still widely used, particularly in the 10G ports shipped globally, however, when the network team wants to provide a backbone in 40G or 100G, they will need to start thinking about fibre cabling, especially when the distances involved exceed 30 meters. This is because of the new Cat. 8 standards can provide 40G over 30 meters. So, it remains a cheaper and viable solution for small distances.”
Fibre-only data centres may not be the most economical solution, but are usually chosen based on other considerations. Consequently, the decision whether enterprise data centres are better off with fibre or copper depends very much on the size, layout and speed requirements and should be studied carefully case by case.
While fibre optic cabling will remain the medium of choice in the data centre backbone, Category 8 copper cabling is certainly a valuable consideration for the data centre edge where uplink speeds in switch-to-server connections are moving from 10 to 25 and 40 Gigabit Ethernet speeds, explains Prem Rodrigues, director for the MEA & India/ SAAR, Siemon.
Automated Infrastructure Management – a must for fibre systems in data centres?
In the data centre world, gradual migration to 100, 200 and 400 Gbit/s is underway. Consequences include higher density and more cabling in racks. Today’s data centres may contain hundreds of thousands of ports and patch cords. The complexity and dynamism of the infrastructures is reaching a scale at which humans can no longer manage them without remote-controlled hardware and software support. What’s more, there are significantly more servers than can be managed on-site by qualified staff.
We are also seeing data centres move towards low loss fibre connectivity, adds Rodrigues of Siemon. “A key driver for this development is the migration of transmission speeds from the current 40 and 100 Gb/s to the next generation 200 Gb/s and 400 Gb/s fibre applications and from 8 Gb/s, to 16 Gb/s and to 32 Gb/s for fibre channel-based SANS and as bandwidth and speeds increase, insertion loss requirements become much more stringent.”
Automated Infrastructure Management (AIM) is an essential tool that can realise benefits for the owner/operator at various locations across their data centre operation. AIM can reduce timeintensive manual processes, minimise human error and network downtime. It can also offer visibility into end-to- end circuits. And, in the event of a network failure, a root cause analysis can be quickly established to help the network to get back online fast.
As higher fibre densities and network virtualisation become prevalent at the core, the challenge of managing those assets also increases. Waite adds that single optical fibre can no longer be associated with a single business process, with virtualised servers the data packets running across a channel may be linked to many disparate elements of a business or even different customers in a cloud type deployment.
Edge data centres can be situated in remote locations, taking a decentralised approach, or placed in urban areas. These edge data centres would be smaller in size, but their numbers would be higher. The edge data centres need to be readily available and accessible for operation and maintenance purposes. Further, 5G will require the support of many edge data centre locations to support low latency applications.
Managing and maintaining such a large number of data centres poses a significant challenge, according to Yousef. “This requires the implementation of new practices on work orders for moves, adds, changes and fault detection. Faster MTTR would be a critical KPI for edge data centres.”
At the edge, AIM can be used to remotely track the utilisation of panels, cabling and switch ports, and provide a real-time view on how physical-layer assets are being used. Such a system also offers numerous extrinsic benefits, thanks to its ability to be combined with other management tools, thus enhancing the security of edge data centres.
In addition to facilitating the management of increasingly large and complex infrastructures, there are other benefits to the use of AIM solutions, explains Khalil. “Using a single, current, consistent database eliminates stranded capacity and facilitates end-to-end analysis, agile infrastructure management, predictive analysis, capacity utilisation and efficiency of operation and administration, and can bring 30 – 50% reduction in downtime. System data can be used for budgeting and IT infrastructure inventory.”
Future steps in AIM include the use of artificial intelligence to ensure infrastructures remain manageable. “AI can independently manage connectivity from the data centre to, for example, a smart city, making predictions based on monitoring and machine learning,” adds Khalil.
New fibre network topologies
The number of fibres used for transport is on the increase. 40 and 100 Gbit/s require eight fibres in parallel pairs. Furthermore, the traditional hierarchical network topology with core, aggregation and access level can no longer cope with today’s needs, resulting in congestion along traffic routes. To ensure that data and applications are available in real-time at all times latencies have to be considerably reduced, calling for singlemode or multi-mode fibre and new kinds of network architecture.
Spine and leaf, supported by a network fabric is the next-generation topology being deployed to replace core, aggregation and access level switching. A spine-leaf architecture reduces latency and can be adapted to continuously changing needs.
Waite adds that this is the most effective way to enable east-west data flow in the data centre. “East-west flow of traffic describes the lateral flow of data across the networking layers within a data centre, allowing servers to quickly and efficiently communicate with one another in support of AI, machine learning, virtual reality and a plethora of other high bandwidth low latency applications.”
A network mesh with crisscrossing cabling guarantees that switches at access level are no longer more than a hop away from each other. All devices are the same number of segments. Unlike the classic three-tier architecture, this new topology has just two layers. The leaf layer is built up of access switches that connect to servers, edge routers, load balancers, firewalls and other devices. The network backbone is provided by the spine layer, which consists of routing switches.
“Meshed leaf-spine architectures, increasing fibre density, complex point-tomultipoint connections, and the fact that devices can be added or reallocated at any time make knowing the exact state of all connectivity elements at all times extremely challenging. Changing network topologies, therefore, also warrants the use of AIM solutions,” says Khalil.
The consequence of deploying this topology is that every cabling channel within the fabric, connecting the spine and leaf switches, must be able to support precisely the same throughput of data. Therefore, each cabling channel must be designed to support the maximum expected data rate; nothing less is acceptable. Commonly the fabric data rate used to be rated at 40GBE amongst many data centre owners. Today this rate no longer applies with many customers only considering 100Gb/s and above as the minimum standard for their fabric cabling going forward.
Arafat Yousef, managing director MEA, Nexans Cabling Solutions.
Nabil Khalil, executive vice president of R& M META.
Prem Rodrigues, director for the MEA & India/ SAAR, Siemon.
Alastair Waite, data centre market development, Commscope.