Elastic Distributed Rendering Service Placement in Capacitated Cloud/Fog Gaming Systems

Abstract

Fog-assisted cloud computing advances have paved the way for new solutions in online gaming, by integrating all rendering resources directly into the edges of the cloud. However, despite the numerous benefits, these systems still remain prone to network delays, ergo, the discovery of an optimal rendering service placement in the fog layer, that effectively reduces capital deployment cost for the game providers and access latency cost for the game clients, is a challenging issue. In fact, it falls into the category of the well-known (NP-hard) facility location problems. As such, conventional centralized approaches, that require global network knowledge, in most cases are deemed unscalable and cost-prohibitive, thus, distributed solutions must be further explored. The approach followed here, introduces a Rendering Service Allocation Policy (RSAP) to autonomously allocate services towards an optimal placement, based strictly on local information available to the fog renderers regarding their aggregate rendering demands, yielding reduced access latency and deployment costs, whilst meeting their stringent capacity bounds. Simulation results showcase the efficiency of RSAP and verify its elastic behavior under different deployment scenarios.