Delay Control for Temporally Fair Scheduling Policies via Opportunistic Mixing

Abstract

We consider the delay control problem for multiuser opportunistic scheduling policies in wireless systems. First, the user selection behaviors of opportunistic scheduling policies are examined to highlight the delay and resource starvation issues. We then propose and analyze two schemes for delay control and starvation prevention that are both effective and flexible for use with existing scheduling policies. The first scheme, the Round-Robin Mixing (RRM) scheme, provides a simple and accurate mechanism for controlling the starvation probability in an i.i.d. environment. Based on the foundation of the RRM scheme and a small set of carefully chosen system approximations, we develop another improved scheme, the Opportunistic Bernoulli Mixing (OBM) scheme, that can support different user channel conditions and resource allocation requirements. The OBM scheme performs much closer in achievable system rate to the original non delay-controlled policy due to its flexible opportunistic delay adjustments while enabling very accurate control on user starvation probabilities. Numerical results illustrate excellent delay/performance tradeoff for the OBM scheme with negligible approximation errors for even moderately small systems.