The Next Internet: A Ubiquitous, Viral Edge Cloud

Опубликовано 14 августа 2017, 14:20

Moving computation and services from the desktop and local back office to large, centralized, multi-tenant data centers ("The Cloud") has been the dominant trend of the last decade in computing. This has led to a vast profusion of new applications and services and broad new areas of research, from networking to building highly-scalable, robust, distributed systems. However, the Cloud is far away from the edge device. Microsoft Azure has been particularly aggressive about building large-scale data centers, but even Azure cannot guarantee better than 30 ms latency. Further, Cloud-edge bandwidth is limited, and for many Cloud applications and services, the dominant cost is not Cloud computing and storage but rather Cloud-edge bandwidth. A ubiquitous edge cloud, with latencies to the edge in the 1-10 millisecond range and bandwidth to the edge in the gigabit range, would usher in a new class of application -- "Cloud-in-the-loop" applications, where the inner loop of the application performs a real-time Cloud transaction. I will give some examples of these applications including a live demonstration of a collaborative visualization application. Building such a ubiquitous edge Cloud using a centralized model demands resources far beyond what is available to a research organization. I will describe an opt-in, viral edge Cloud where individuals and institutions offer local VMs on an as-available basis to the edge Cloud, and its early prototypes. I will also describe a programming environment for wide-area distributed systems that can rest on top of this (or any) Cloud, which will permit the assembly of highly distributed applications and services from existing, persistent services.

See more on this video at microsoft.com/en-us/research/v...