The idea of finding a theory of everything providing the minimal set of abstractions for dealing with data in motion, data at rest and computations, at any scale, has always fascinated and to some extent obsessed us. We have been working on this problem for years, refined our abstractions, evolved our sensibility, and deepened our understanding of the problem — after years of sweat, we think we have possibly found the answer.
The answer to the problems that those of us involved in the design of distributed systems have struggled-with over several years. The problem created by the segregation and impedance mismatch between technologies for dealing with data in motion, technologies for dealing with data at rest and those for dealing with remote computations.
In this presentation we will introduce the four abstractions that we have found able to be necessary and sufficient to express data in motion, data at rest and computation along with the handful of primitive to operate over them. We will Illustrate how these four abstractions have been efficiently implemented within the Eclipse Zenoh project and how they can be composed to build a data management framework that is able to support any-scale data in motion, geo-distributed storages and queries along with distributed computations. All of this, while providing location and referential transparency.
We will also provide examples of how Eclipse Zenoh is being at the edge-computing data fabric in several application domains such as Robotics, Autonomous Vehicles, 5G, and Smart-*.