Project dates: 01. May 2020 - 30. Apr 2022
Objective
In an era that Internet of Things (IoT) applications are getting increasingly dynamic and mobile, the Achilles heel of such efforts remains the management of the application state (e.g., consistency, availability, performance, fault tolerance). For stateless applications, functionalities like load balancing (active-active), failover (active-passive), and job continuity (roaming) are easily addressable at the network level. For stateful applications though, the developers must provision a clustered data layer for managing the ever-growing collections of geographically spread data. This layer needs to provide geo-transparent data access, integrate well with the interplay of backend services, and be tailored (described by structure and purpose and expressed in functionality) to the specific application requirements.
Over the last few years, research efforts are mainly focused on security, privacy, confidentiality, energy efficiency, and reliability of storage systems. An important aspect that has not been thoroughly investigated yet is programmability. To keep pace with the increasing birthday of new application models it is essential to pass the technological barrier of 'programming storage systems' and enter the age of 'composing storage systems'. So far, no competent solution has been proposed either by industry or academia.
Herein, we propose an open-source development ecosystem intended for engineers and researchers looking to invent, build, verify, evaluate, and release storage systems, tailored on a per-application basis. The ecosystem, called Ether, consists of i) the virtualized storage platform for managing application state whose data are spread across heterogeneous services provided by third-party vendors ii) the virtualization framework for specifying the control flows and the data management of the platform, as hierarchies of virtual functions iii) and the community hub for sharing virtual functions, platform configurations, or best practises.