Yarob Sakhnini, Systems Engineering Manager, CEMA at Brocade Communications
When it rains, it pours. The dying years of the last decade witnessed a recession that was almost without precedent in its severity, and certainly unequalled in its worldwide reach; at exactly the same time, the juggernaut of the information revolution ploughed relentlessly onwards, throwing new IT challenges at organisations across the globe.
Chief among these challenges was the wholesale migration of computing into the Cloud. As businesses of all sizes shifted their IT function off-site, so the strain increased on the world’s data centres. Added to this was the deluge of data as organisations created, used and stored ever-more digital information. Together, these created a perfect storm, at the eye of which sat the data centre.
The more that organisations in the Middle East have come to depend on the data centre, so the demands upon them have grown. As more business-critical data and documentation is held in storage facilities, and as we have come to rely on the cloud for more and more of our day-to-day computing needs, so the threads of the data centre network fabric are increasingly under pressure.
The situation presents a double dilemma – organisations, under pressure to cut operating expenditure, seek solace in the Cloud’s promises of lower costs; and this places greater strain on traditional data centre networks that were designed for a gentler, less information-intensive age.
Yarob Sakhnini, Systems Engineering Manager, CEMA at Brocade Communications says that Virtualisation has been hailed as the answer to these twin problems, and with good reason. The ingenuity of maximising hardware usage by creating “virtual machines” has, at a stroke, solved many of the problems associated with the data and cloud computing boom – not least by abolishing the unacceptably wasteful and costly practice of having bank upon bank of under-utilised servers.
Virtualisation has been around long enough for most organisations to understand its importance as the engine that powers the Cloud and enables data centres to provide such a wide range of storage, processing, infrastructure and software services at a low cost. Less well known, however, is that virtualisation is not a simple panacea for our addiction to data and cloud computing.
Virtualisation is not, as some may think, simply a case of adding more capacity by cramming more virtual machines onto every physical server. In fact, this proliferation of virtualised servers actually adds to the complexity of the wider data centre environment by increasing the number of logical connections vis-a-vis physical connections across the network. The bandwidth available to the data centre must therefore be split between an even greater number of independent channels, each of which can be assigned to any server or device at any time.
Traditional networking technology – the switches, cabling, routers and other “bolts” that make up data centre infrastructure – simply cannot cope effectively with the increased traffic volumes and the diverse set of applications of virtualised data centres. Furthermore, high levels of server utilisation places even greater demand on the network to ensure bandwidth availability which, given that many organisations entrust the greater part of their storage, applications and other vital functions to the cloud, must be maximised at all costs.
The key to creating a successful virtualised datacentre therefore lies not in the choice of hypervisor or virtualisation technology, but in modernising the very infrastructure that stitches the network together – especially if this infrastructure is based on legacy Ethernet.
Over the decades Ethernet has been a trusty servant to IT networks, and it is a technology that has constantly evolved as new types of architecture have emerged. The latest evolution, Ethernet fabric, is a revolutionary networking mantra that is perfectly suited to the demands of the next generation of data centres powering the Cloud, and which overcome the deficiencies of legacy Ethernet technology.
Classic Ethernet networks are hierarchical with three or more tiers. Traffic has to move up and down a logical tree to flow between server racks, adding latency and creating congestion between network switches. What’s more, legacy Ethernet networks require switch management, where each switch and port has to be individually configured; as more servers are added, so the number of switches increases. One of the key limitations of this model is that it relies on Spanning Tree Protocol (STP), which prevents loops within the network by only allowing a single active path between any two switches. This limits the amount of bandwidth to a single connection, which is singularly unsuited to the new traffic patterns necessary for virtualised environments.
Ethernet fabrics, by comparison, creates a flat, multipath network where every switch knows about each other, all paths are available and traffic automatically travels along the shortest path. This ensures maximum bandwidth availability, lower energy consumption, fewer adapters and switches, and better transfer of data packets across the whole environment. In short – and a full discussion would take many pages – Ethernet fabric is a more intelligent and holistic “stitching together” of the network infrastructure.
Ethernet fabric is the sine qua non of a successful cloud environment, but it is not a single magic bullet for successfully implementing a virtualised environment. That can only be achieved by working with vendors to deploy dynamic data centre infrastructures that are suited for their particular demands – and which, naturally, are based on best-of-breed Ethernet fabric technology.