The demands being placed on data centres are growing bigger all the time. Expectations of faster speeds and higher bandwidth mean that all data centre managers need to be looking at how they can transform their existing operations in order to cope with this. But what will they need to consider to make a success of such a move?
A recent white paper by CommScope noted that current trends suggest data centre bandwidth demands are set to continue growing at between 25 per cent and 35 per cent a year. One of the key drivers for this is the shift to higher switching speeds.
In the past, one of the common setups for a data centre has been the use of 40Gb ports, each supplying four 10Gb server connections. However, as data demands have increased, these solutions have become overwhelmed.
Therefore, the data centre industry is quickly moving to 25Gb and 50Gb alternatives. But with lane capacities expected to continue doubling, 100Gb ports are likely to be commonplace by around 2020. At the moment, many within the industry are moving to 25Gb lanes as their next switching technology, as this offers easy migration to 50G, 100G and beyond as technology evolves, thereby representing a much better long-term return on investment compared with 40Gb solutions.
CommScope's white paper identified a number of factors that are contributing to the increase in data centre throughput speeds. For instance, it noted server densities are rising by around 20 per cent a year, while virtualisation density is growing by 30 per cent. There has also been a profound shift in the profile of traffic within data centres, with east-west traffic now far surpassing the volume of north-south traffic.
"Network design has to reflect this massive amount of traffic, and, importantly, has to allow for server, storage, and network capacity to all be scaled up independently and with as little disruption and reconfiguration as possible," the white paper stated. This means data centre professionals must be able to support higher server densities, deploy more fibre and accelerate plans to migrate to higher speeds in their core and aggregation networks.
A central part of any high-speed data centre migration will be adapting the architecture to better accommodate east-west traffic. Traditionally, data centre architecture uses a three-layer approach, with a core layer feeding into an aggregation layer that connects access switches and an access layer that runs from the data centre to the individual nodes where users connect to the network.
But while this setup is designed to handle mainly north-south traffic, it is not the most efficient way of coping with east-west traffic.
As a result, there has been a significant shift recently towards a 'spine-and-leaf' architecture. This means networks are spread across multiple leaf switches, with each leaf switch connected to every spine switch, thereby creating a highly resilient any-to-any infrastructure.
However, such an approach will require a greater number of fibre connections. CommScope said: "Given the increasing density, cabling connectivity and management solutions such as optical distribution frames, panels and raceways become more important."