ComputerWeekly asked BridgeWorks CEO, David Trossell, to talk about data processed at the data centers edge, rather than the core will spell the end of a centralised cloud for IoT, mobile and devices.
July 19, 2017
How cloud, IoT and mobile devices are driving demand for edge data centers
Enterprise appetite for micro and edge computing is one of the top 10 trends affecting the industry, according to analyst house Gartner, with organisations increasingly looking to place datacentre resources closer to users and their devices.
But what is driving these trends and what does it mean for enterprise datacentre and infrastructure strategies?
There are three motivations for using edge datacentres and cloud resources. The first and most obvious is convenience, according to Jack Bedell-Pearce, managing director at UK datacentre, cloud and connectivity provider 4D.
“If an IoT [internet of things] organisation deploys a large number of devices in a new region, if they want to be able to provide support or need to scale quickly, having a datacentre in that region will save engineers a lot of time and allow them to respond to issues – such as minimising downtime – much more quickly,” he says.
Cost is another important factor, because running and maintaining a city centre datacentre is notably more expensive than doing so in areas where rent and business rates are not so high. “If strategically deployed, companies can make significant cost savings by distributing their infrastructure in regional datacentres,” adds Bedell-Pearce.
Lower latency is also another benefit of using edge facilities. After all, pushing computation and analytical capabilities closer to the network edge reduces traffic over the network. This in turn can greatly reduce the round-trip delay in sending data for analysis in a centralised cloud platform, offering additional privacy, resiliency and security.
That’s according to Rhonda Ascierto, research director for the datacentre technologies and eco-efficient IT channel at market watcher 451 Research.
“IoT gateways will communicate with local datacentre aggregation sites for data pre-processing and/or real-time analytics such as error detection of security monitoring, as well as optimisation for transmission upstream to datacenters in the core (non-edge) layer. Automated vehicle navigation is one such examples,” she says.
Are edge datacentres anti-cloud?
It could be argued the trend towards edge datacenters is the antithesis of cloud, given that concept is all about pulling data to the centre, rather than pushing it towards where users are. Jessica Storer-Woods, product manager at cloud managed services provider (MSP) Timico, disagrees, and states there is still a massive demand for centralised cloud resources. “Line of business applications, operational functions and legacy systems that need to sit in the core, protected by the security that the core brings,” she cites as examples.
Instead of considering them to be the opposite of cloud, edge datacentres should be considered an enabler and extension of it.
“Take the ‘distributed cloud’ notion, which is essentially a federated approach, supported by networking infrastructure such as Wi-Fi and eventually 5G, for workloads to be housed and moved across the (fiber and cellular) edge, and elsewhere,” she says. Next-generation networks, such as 5G, are expected to further enable this trend, she adds, making distributed cloud more than just a replacement or upgrade for telecoms central offices.
With many IoT devices deployed in the field and in areas of low connectivity, it is important their access to data is quick and reliable, which is further driving demand for edge datacentres. “Businesses using IoT will be early adopters and therefore need to have as much confidence as possible that their access to data and connectivity is as quick as possible,” says Storer-Woods.
Ascierto says the edge of the network is under strain from the growth in high-speed connections and traffic volume. “These can clog access points and the network switches – unless these are upgraded to handle many more concurrent connections and higher data rates,” she says. “The shift to 1 gigabit per second switch port speeds is ongoing as wireless devices flood the networks, streaming video, music and other content. “The adoption of software-defined networking and commodity off-the-shelf [Cots] hardware will increasingly drive the use of standard IT equipment for edge switching,” Ascierto adds.
The end of the cloud?
Does this move back to having data processed at the edge rather than the core spell the end of a centralised cloud? David Trossell, CEO of Bridgeworks, does not think so, adding there will always be a need for core datacentres.
“While customers are often pushed to buy into the latest and greatest technology, what they really need is choice. Sometimes it just doesn’t solve the issue they wish to resolve, or meet their needs,” he says.
That said, edge computing may be too expensive for some organisations right now, while the cloud offers scalability and economies of scale. “Organisations need to clearly define what they need, and be sure that edge computing is the most viable option for them over other options that are available on the market,” he says.
Location, location, location
Where best to locate an edge datacentre is largely workload-dependent, says Ascierto, because capacity may be required at both the edge – where devices are – and the near-edge.
IoT gateways aggregate sensor and smart object data streams and provide a critical networking layer enabling, among other things, local-area device data to connect directly to broader networks, such as a WAN or the internet, for delivery to local (edge) or centralised datacentres or, increasingly, provide intelligence edge computing capabilities.
“The myriad examples inherent in IoT mean that there will be both inexpensive sensors and gateways that tunnel traffic back to private or public cloud services for ex-post facto analysis, and also more mission-critical latency-sensitive applications that require localised database and analysis,” she says.
The move to the edge is not something that will lead to a completely distributed model with no core computation. There will still be core datacentres, as there is very much a need for them. But hybrid models that process and store data wherever it is more appropriate to do so are likely to arise over the next few years.
With 5G, there is also the prospect of those base stations containing micro-datacentres that bring the processing and streaming of data and the edge even closer to the user.
