Autonomic operation of massively multiplayer online games in clouds

Abstract

To support the variable load of Massively Multiplayer On-line Games (MMOGs) with millions of registered users and thousands of active concurrent players, game operators over-provision a large static infrastructure capable of sustaining the peak load with guaranteed Quality of Service (QoS). This leads to inefficient resource utilisation, high service prices, and limited market participation accessible only to the large companies. To address this problem, we propose a new autonomic ecosystem for hosting and operating MMOGs based on cloud computing principles involving four smaller and better focused business actors whose interaction is regulated through Service Level Agreements (SLAs): resource provider, game operator, game provider, and client. In our model, game providers acquire operation SLAs from game operators to satisfy client requests and manage multiple distributed MMOG sessions. Game operators lease on-demand cloud resources based on the dynamic MMOG load and guarantee the required QoS to all clients. We evaluate through simulations based on real MMOG traces and commercial cloud SLAs different methods of ranking MMOG operation offers. We show that considering compensations for SLA faults in the offer selection can lead to over 11% gains in game providers' income, and that adequate ranking of offers can reduce operational costs by up to 60%.