Resource Allocation for Semi-Elastic Applications With Outage Constraints in Cellular Networks

We consider resource allocation for semi-elastic applications, such as mobile video conferencing, which require a maximum outage. We represent the performance of a session by a sigmoid utility function of the average bit rate over a time window consisting of many slots. The goal is to maximize the total expected utility of all active users. A principal challenge is the difference in timescales: The outage is measured over sessions, whereas the average bit rate is measured over a time window such as a group of pictures. We propose that these differences in timescales can be elegantly addressed by shadow prices: a price per unit average rate over each time window and a price per unit outage. We further propose that the price per unit average rate depends on combined path loss and shadowing but not on fast fading. We show that resources can be efficiently allocated if the base station chooses prices based on the total demand and if the users respond by choosing average rates. The performance of our algorithm is illustrated by simulation results.