Hardware Implementation and Analysis of MAC Schedulers for Wireless Networks
Today’s wireless networks have the potential to support different kinds of applications like online gaming, live video streaming, v2x communications, etc. Apart from obvious throughput requirements, many of these applications require strict delay constraints on endto-end experienced by an application packet. Medium access control (MAC) layer is an essential part of modern wireless technologies and is responsible for radio resource allocation. Quality of service (QoS) performance of a user depends heavily on the scheduling decisions by the MAC layer. For example, a user with a tight delay budget should be given priority in resource allocation. To the best of our knowledge, popular scheduling algorithms like proportional fair and round-robin do not consider these constraints as a parameter in their scheduling rule. Head of line (HoL) delay, which is the sojourn delay of longest waiting packets in the queue, can be an important parameter to perform scheduling for data traffic with delay constraints. We implement delay sensitive proportional fair scheduler (DS-PF) and exponential delay sensitive proportional fair scheduler (Exp DS-PF) scheduler based on modified largest weighted delay first (LWDF) scheduling rule in LTE equipment setup in the department. These scheduling algorithm takes into account HoL delay, packet delay budget, and shortterm throughput in scheduling rule. We compare its performance with popular scheduling algorithms like proportional fair. Moreover, to implement DS- PF scheduler for up-link scheduling, we implement an HoL estimation algorithm for the up-link scheduler. HoL information for traffic flows or queues is available at MAC layer in down-link as these queues are maintained at eNB itself. This HoL information is not available at the MAC schedulers in up-link because packet queues are maintained at user equipment (UE) and it would take a lot of resources to transmit this information to eNB. Implemented HoL estimation algorithm utilizes the buffer status reports (BSR) containing the volume of data user equipment needs to transmit. Information from BSR and time intervals between multiple BSR is stored in virtual queues at eNB to estimate HoL. The size of the virtual queue is taken according to the delay budget of the respective traffic queue or flow. For the comparison of the performance of the DS-PF scheduler using estimated HoL values, we implement load sensitive proportional fair (LS-PF) scheduler based on the modified largest weighted work first (LWWF) scheduling rule. LWWF scheduling rule uses work information, which basically means the amount of data that the UE has to transmit in its scheduling rule. This information is easily available in the up-link at MAC layer. Therefore, LS-PF serves as a critical scheduler to signify the necessity of estimating HoL values in the up-link scheduler.
From different experiments on implemented scheduler, it is shown that the DS-PF and Exp DS-PF scheduling algorithms perform significantly better in terms of end-to-end packet delays for user equipment containing delay sensitive traffic flows. This concludes that the schedulers considering HoL delay and delay constraints in their metric, perform well for traffics flows with QoS constraints (packet delay budget). For up-link scheduling, the scheduler utilizing the estimated HoL value performs better in terms of end-to-end packet delays than the LS-PF scheduler for user equipment having delay constrained applications. This confirms the real-world advantage of estimating HoL in up-link scheduling to improve the QoS performance of the network