Re-designing Cloud Platforms for Massive Scale Using a P2P Architecture

Abstract

Cloud platforms need to scale with the number of resources and users they manage, while maintaining the needed performance levels with respect to service parameters such as application deployment time, service availability and response time. With the increase in capacity of todays data centers and distributed cloud deployment scenarios like edge computing, the scalability requirements of a cloud management software has become paramount. Most commercially available cloud management software are centralized from a software architecture point of view, and this often poses a limit on the number of resources that the software can manage while maintaining an acceptable performance level. In this paper, we present a generic design based on peer-to-peer (P2P) architecture for scaling cloud management software. We demonstrate how this design can be realized by applying it to the OpenStack cloud management platform. Our implementation alters neither the core functionality of OpenStack nor the way users interact with OpenStack. Performance evaluation results of the implementation show that, when used in a large system under high load, the solution greatly improves VM boot up times and VM startup success rates.