Demand for edge computing is growing rapidly. Deloitte, for example, predicts the market for edge technology will reach $12bn (£9bn) this year. Other surveys suggest it could be worth more than $60b by 2028.
That growth is being driven by the internet of things (IoT), 5G connectivity and distributed systems in fields as diverse as manufacturing and healthcare, remote surveillance and even by developments such as autonomous vehicles.
Forrester, for example, talks about the “four edges of edge computing”. These comprise: the engagement edge, where consumers or the environment interact with the device; the operations edge; the enterprise edge; and the provider edge. This is a long way from the situation even a decade ago, when edge computing meant small or branch office IT and perhaps a small number of enterprise-owned mobile devices.
Edge systems are helping organisations respond to growing data volumes, and the need to process and analyse that data, without the cost and latency that comes with transferring information streams to a central datacentre.
Edge systems, however, need edge storage. And that storage needs to be adapted to the workload, location and environment.
Where is edge storage deployed?
Edge infrastructure and edge storage is by no means new. Organisations ran IT in remote and branch offices (ROBO) even before the advent of the desktop PC. The move to desktop PCs was quickly followed by local servers or network-attached storage (NAS) devices to provide file sharing and backup. More technically mature organisations used local storage to stage backups for uploading to a datacentre or to offload to tape.
User devices can also be viewed as edge computing, and in some ways pose an even greater challenge than a ROBO setup. The increasingly mobile nature of personal computing makes managing storage, including backups, even harder than backing up branch office hardware. The cloud has at least eased some of the pressure, allowing automated backups directly from a personal device. But organisations still need to manage local device storage, especially when it comes to security.
Increasingly, though, edge storage will be deployed in remote and autonomous devices. As well as sensors and surveillance hardware, these are likely to include vehicles, traffic and other environmental management systems, and, increasingly, 5G base stations.
All of these require at least some local storage.
Why use edge storage?
Enterprise IT departments are – rightly – focused on storage in the datacentre, or in cloud infrastructure. These are large, centrally managed, scalable systems.
Moving data from the edge to the centre, however, is not always possible or desirable. The vast volumes of data that originate from connected sensors, systems such as CCTV, and other IoT technology could quickly overwhelm network connectivity.
Even if that wasn’t the case, latency, quality of service and reliability are all arguments against centralising all data. Processing locally and only sending a portion of significant data – perhaps results, exception or fault data, or even time-based data samples – reduces demand on bandwidth. But it needs efficient local storage.
“Over recent years, we have witnessed the explosion in IoT – internet-connected devices – which has been the main driver behind edge computing systems,” says Toby Alcock, chief technology officer (CTO) at IT integrator Logicalis. “With more devices connected to the internet that require real-time computing power, edge computing has enabled data storage to be closer to where it’s gathered.”
A further variant on edge storage, and one being explored by the telecoms industry, is edge caching. Here, data is stored on a temporary basis before being moved on to other, centralised or cloud, systems for processing.
Edge caching can be used to apply data compression, encryption, or tasks such as load balancing. Edge caching is already in use in the media and internet industries in content delivery networks (CDNs). Unlike other edge storage applications, caching is used to hold data being delivered downstream to the consumer, as well as upstream to the enterprise.
Edge storage: What technology?
From a technical perspective, edge storage is most likely to be solid state. Computer-based systems are likely to use solid-state disks (SSDs) or non-volatile memory express (NVMe). Larger IoT devices use NVMe, but surveillance cameras and other low-power equipment will use removable storage, often in the form of a memory card.
MicroSD cards are a popular choice among manufacturers, with capacities now running at 1TB (terabyte) and above and SD card specifications now supporting up to 128TB. Manufacturers are also producing enterprise-grade SD and microSD cards, and types optimised for video surveillance and other applications that require constant writing of data. These cards are more robust than consumer-grade cards, but much more expensive.
Edge storage also needs to be physically robust. IoT devices might be exposed to the elements or fluctuations in temperature, and they might be subject to vandalism or theft. In applications such as the automotive and transport sectors, media needs to be able to handle vibration and motion.
The edge storage device needs to work without the protections available in a datacentre or even an IT rack. “Edge technology needs to fit the form and function of the infrastructure,” says Forrester analyst Naveem Chhabra. “Think about the example of autonomous cars. You can’t put a full rack server in a car. You don’t have the space or the cooling.” At the engagement edge, an interactive device such as a retail kiosk has its own design constraints.
Edge storage pros and cons
Installing storage adds to the cost of edge devices. The justification is that storing (and processing) data locally is cheaper or more efficient than alternatives such as transferring to a datacentre or the cloud.
The cost of bandwidth is the key factor driving storage to the edge. Industry experts suggest that based on Amazon Web Services (AWS) pricing, bandwidth costs around four times as much per gigabyte as storage. Stripping out redundant data before it is uploaded saves costs.
The other factor is the difficulty of reaching the edge with conventional networking technologies, and the reliability and latency issues that can affect networks over distance.
“One of the big challenges – or perhaps I should say big motivators – for any edge technology is latency and bandwidth. That is, the speed and capacity of the connection back to the core,” says Bryan Betts, an analyst at Freeform Dynamics. “If you’re trying to minimise the data you send and react faster, which you normally will be, you inevitably need to do more at the edge.”
The challenge is that edge storage, because of its location, is harder to manage and maintain. Upgrading capacity will mean a hardware swap-out, and offloading data relies on those limited bandwidth connections into remote locations.
Nonetheless, suppliers and customers are looking at ways to improve edge storage capacity, particularly by offloading data, either over network during quiet times, or even physically.
“Edge storage doesn’t scale very quickly compared to the cloud,” says Ivaylo Vrabchev, head of professional services and an AWS solution architect at HeleCloud. “Accurate capacity management is needed to protect the system from running out of resources.”
