3 reasons why Hyperscale is the new driver of Data Centres
- 20 April, 2017 14:00
The swarm of small data packets (and the signalling that goes with it) can dwarf large data transfers on a network. (Picture: Niv Singer, Flickr)
The proliferation of the Internet of Things (IoT) is putting pressure on data centre infrastructures like never before. With the number of connected devices expected to reach 21 billion by 2020, we’re on the brink of a major technological revolution. Carrying the potential to reshape every facet of the enterprise, from business models to technology, data has become the backbone of our evolving digital world. Organisations are quickly realising that their data strategy is, in fact, their business strategy.
We’re seeing a rapid expansion of cloud services to support increasing data storage requirements. And as data becomes more and more critical to success, so too does the ability of the data centre in supporting the growth of cloud solutions. The complexity of new workloads and increasing demand from customers is leading to different server, network, and IT requirements in today’s data centre.
Welcome to the era of the hyperscale data centre.
The Cisco Global Cloud Index predicts the number of hyperscale data centres will grow to 485 by the end of 2020, after reaching 300 in the US at the end of last year. Locally, Australia is proving to be a location of increasing activity for the hyperscale data centre: Amazon Web Services has two data centres in Sydney, and Microsoft also has data centres in Australia.
With this anticipated growth in mind, let’s take a look at why hyperscale is becoming the new engine for data centres:
1. Complexity and scale of cloud applications is rising
Migration to the cloud is the number one driver behind the increasing demand for the hyperscale data centre. The importance of agile development and new application workflows to drive innovation is second to none. This is challenging for banks, retailers, and even law firms – their expertise doesn’t lie in managing a building full of servers.
The availability of cloud-based services provides quick and easy access to the platforms needed to support innovation because of its dramatically reduced barriers to entry: with an Amazon Web Services account, new projects are set up in one day and operate on a usage-based consumption model. If a company’s website is booming, they can easily fire up another 50 servers. The hyperscale data centre is enabling this, due to its ability to deliver large volumes of product very quickly. It can also scale up capacity in smaller increments. With the correct optical cabling and technology, a hyperscale infrastructure supports migration to 100GB speeds and beyond.
2. Demand for a seamless customer experience
All organisations are looking to leverage the data they have in order to gain customer insights, the goal being to provide better service outcomes for their customers. A high-performing hyperscale data centre is key to doing just that. These data centres rely almost exclusively on fibre optic cabling for interconnectivity between servers, storage, and switching. Looking forward, we’re only going to see an increase in the number of organisations transitioning to the cloud, as they look to derive the outcomes they want faster. Without optical cabling, data centres would be unable to evolve to meet future capacity requirements. Optical cabling affords larger data transmission rates over longer distances. This is essential to supporting the enormous amounts of data that organisations transport, in increasing quantities, within their data centre environments.
3. Shift in data centre architecture
The fact that the IoT is generating rapid waves of data bodes well for the data centre industry in general, and specifically for the hyperscale segment. Companies that make and sell connected devices are seeking cloud solutions to manage data. As the IoT drives the shift towards the hyperscale data centre, it’s changing the physical layer of the network.
Where a data centre traditionally leveraged the 3-tier switching model, we’re seeing hyperscales move towards a 2-tier spine-and-leaf architecture. Spine-and-leaf significantly reduces latency when it comes to accessing data, as it moves data across the physical links in the network. While flatter, this results in more physical cabling in the data centre itself, as every spine switch needs to be connected to every leaf switch. The hyperscale model is ideal in supporting the deployment of more cables, as it offers higher density. This is increasingly the networking architecture of choice for hyperscale cloud providers.
As the explosion of data continues to increase, the more important it becomes for the data centre to meet future capacity requirements. At the same time, customer expectations for a seamless experience have never been higher. The hyperscale model provides a massively scalable, future-ready infrastructure that enables fast transmission speeds and seamless migration.
And as organisations continue to go global with cloud storage in data centres, we’re seeing hyperscale[MLS1] s expand from its US roots and become more of an international game. The next game changer will be Chinese operators making an international mark and attempting to take on the likes of Amazon Web Services and Google. Watch this space.
Clive Hogg is Technical Manager, Corning Optical Communications