The model for an edge data centre is a small, compact Machine room with a high density of Compute and good network conductivity. The objective of these data centres is to provide local datasets to local users, which significantly reduces latency as there is no requirement to pull data from geographically distant mega data centres. The concept of an edge data centre is not new, Telco’s have been using caching edge proxy servers to cash relevant data needed for 3G and 4G data sets for a long time. These servers reside within the mobile data networks, normally close to the mobile phone masts and are used to cache data needed by smart devices locally to the network. They speed up data provision and reduce network latency as well as reducing overall data loads on the Core network.
Edge data centres are a natural progression of caching edge servers and the need has been driven by the growth in Digital and the corresponding growth in data. In the modern smart urban world, people are increasingly reliant on their smart apps to support their day-to-day life. Be this hailing a smart taxi, planning the most effective route to work or utilising urban services through smart devices. All of these functions require compute to deliver them. Although the urban smart app only requires a subset of data, this data must be readily available and local in order to ensure users are not sitting there watching the hourglass.
The Edge data centre evangelists claim that all datasets shall reside in the edge at some point in the future. Do we believe this to be a correct statement? Like all things it depends. Anyone with a memory of the development of Compute knows but not all applications need the same kind of Compute capabilities. Some applications for example real-time transactional data; Big data analytics; or Database applications tend to prefer large centralised compute environments and work better in mega data centres. If the role of Edge Data centres does grow, it will only occur when the applications are architected to run in a fully distributed environment similar to a grid. With the explosion of cloud, we are seeing grid like applications being developed, but the grid is held within the mega data centre. For these workloads to transition to Edge Data Centres the architecture needs to become fully distributed with compute nodes found within separate edge data centres and network traffic can reliably distribute the data.
So in summary for the Edge Data Centres to take over from existing Mega Data Centres, we need application and solution architects to be able to architecture the applications to function in these environments and we require network performance to at least meet the network performance which exists within current mega data centres. When this happens the role of Edge Data Centres will become more prevalent. There will always be workloads which do not meet this criterion and we should not expect our efficient Mega Data Centres to disappear any time soon. They are amazingly cost-effective at delivering compute by delivering economies of scale through size.
We should just look at the main frame, its demise has been predicted for decades, yet it is still sitting at the core of so many applications and shows no sign of going away anytime soon.