The Internet Of Things will be depend on fast, local compute and storage built on highly distributed cloud platforms.
Traditional Cloud Models Won’t Cut It
Traditionally, the cloud has been construed in much the same way as on-premise infrastructure: a centralized source of compute and storage. The cloud is more flexible, more agile, and often less expensive than traditional infrastructure, impacting how businesses think about IT management and utilization. But for the most part, cloud infrastructure has been centralized infrastructure.
That’s ideal for many use-cases, especially those that leverage the cloud for data aggregation, storage, and analysis. The distribution of data has focused on redundancy and replication. Distribution for performance has been limited to replication across relatively wide areas, often on the scale of continents. And that has been enough for many applications, but for the Internet Of Things, particularly that part of the IoT referred to as the Industrial Internet, such high-level distribution probably won’t be sufficient.
Putting Data Close To Its Zone Of Influence
There are billions of sensor-equipped devices connected to the Internet. Many are consumer devices like smart phones. But one of the biggest areas of impact will be industry. Because of the high-fidelity view of the world the IoT can provide, business decision making-processes will be fundamentally changed, and in many cases automated.
The Internet Of Things will depend on the ability to aggregate, analyze, and react to data on very short timeframes — often in real time. To achieve that, compute and storage will need to be placed in proximity to devices — in some cases that will mean on-premise infrastructure, but in most circumstances the best option will be a highly distributed multi-cloud platform.
A Three Tier Cloud
Several models have been proposed for how the cloud can accommodate the Internet Of Things and its demands for distributed infrastructure. One of the most realistic is a three tier model:
The Fog – A significant proportion of IoT data processing will take place at the edge of the network — on the devices themselves or very close by. Devices will be able to communicate with each other to share data.
Distributed Cloud – While data can be usefully processed at the edge, there are many situations in which data aggregation is essential. For example, data from sensor-equipped trucks can be aggregated to analyze vehicle performance across a fleet to optimize delivery schedules in real time. This sort of data is not hyper-local or appropriate for edge processing, but for optimal performance and responsiveness, it is better handled “close to the ground”. This is the perfect application for geographically dispersed cloud platforms.
Centralized Cloud – This tier reflects typical cloud use today. Centralized data lakes and large-scale analytics processing will still be essential, but will work on longer time frames than edge and distributed processing.
Some have seen these tiers not as part of an integrated whole, but as fragmentary and harmful to the cloud ecosystem. I don’t think that’s the case. Technology like that provided by ComputeNext’s cloud integration layer — which can be used to build both centralized and geographically distributed cloud environments — will empower IoT platform builders to avoid the potential complexity and provision the infrastructure they need, where and when they need it, without the management headache.