Enhancing coordination in cloud infrastructures with an extendable coordination service

Abstract

With application processes being distributed across a large number of nodes, coordination is a crucial but inherently difficult task in cloud environments. Coordination middleware systems like Chubby and ZooKeeper approach this problem by providing mechanisms for basic coordination tasks (e. g., leader election) and means to implement common data structures used for coordination (e. g., distributed queues). However, as such complex abstractions still need to be implemented as part of the distributed application, reusability is limited and the performance overhead may be significant.In this paper, we address these problems by proposing an extendable coordination service that allows complex abstractions to be implemented on the server side. To enhance the functionality of our coordination service, programmers are able to dynamically register high-level extensions that comprise a sequence of low-level operations offered by the standard coordination service API. Our evaluation results show that extension-based implementations of common data structures and services offer significantly better performance and scalability than their state-of-the-art counterparts.