"The end of cloud computing is right around the corner." What? We were just getting started.
In this talk from Peter Levine of Andreessen-Horowitz, he looks forward and declares an end to cloud-computing.
Actually his talk is part of a larger trend, data is increasingly generated by sensors (Autos, Airplanes, Drone, Ships, Shoes) and these devices are generating such large amounts of data in real time such that sending it to the cloud for decisions is not feasible. Essentially autos would crash due to the round trip latency to analyze a partially obscured stop sign.
History says we repeat the patterns of consolidate and then distribute. Mainframes to client server, dedicated servers to virtual machines, stand alone applications to web applications. Today, we are in the consolidation phase with cloud but another wave of data is going to be generated that will require us to decentralize again.
This shift of decentralization is due to endpoints (or devices) having to make real-time decisions on the capture of large amounts of data. This means the endpoint will have to be capable of processing data in real-time and making decisions locally through machine learning processing. Building the machine learning algorithms will still require a centralized computing platform where endpoints are smart enough to only send the data required to enhance algorithms.
Today, we are already seeing this shift of endpoint computing in autos where small and powerful computers are driving. Imagine 192 supercomputers in a small form factor, you now have a rolling datacenter.
Lastly as each of these consolidate and distribute waves have occurfed, the order of magnitude of devices will geometrically rise. Moving to edge based computing means trillions of devices and endpoints with exabytes of data.
Why do we care at Igneous?
We saw the ever growing amount of sensor data (large unstructured data) that was not being served by current infrastructure. These sensors in autos, airplanes, media, and bioInformatics where generating petabytes of data and moving petabytes of data to the cloud would be a real challenge. Data needed to be captured at the creation point and processed in place - we called this data-centric computing. In the future, moving more compute to the edge just makes sense with the cost and fidelity of sensors continuing to decrease.
We even have our own small and powerful computer for processing data - today this lives on premises as part of the Igneous Data Service. Tomorrow it may live closer to the edge.