Speculative resource allocation for packet-switched wireless networks

In this paper we study benefits of speculative scheduling in wireless networks with elastic services. Common goals of Radio Resource Management (RRM) have traditionally been maximisation of throughput and provisioning of QoS guarantees. Providers of wireless access, however, need to acquire adequate revenue to sustain their business. While throughput maximisation and QoS guarantees increase the quantity or quality, respectively, of the chargeable goods, it has also been shown that by accounting for users' perceptions of what is acceptable from QoS and pricing perspective, provider revenue may increase. We propose and evaluate a packet scheduler for realtime streaming services with the aim to maximise user satisfaction and, thus, improve operator revenue. We also extend an existing service acceptance probability model with a service continuation probability model. The proposed scheduling scheme is compared to a proportional fair scheduler. Preliminary results show that revenue can be increased significantly, while keeping user satisfaction approximately constant. Alternatively, the user satisfaction can be correspondingly improved for fixed revenue.