Analysis and optimization of the (E)GPRS RLC protocol by simulations and measurements

A detailed analysis of the GPRS and EGPRS RLC/MAC protocol and its influence on the overall end-to-end performance has been performed. The end user application throughput depends on several parameters, such as the RLC polling period, the RLC window size, the RLC roundtrip time, and specific protocol implementation issues. Simulation and measurement results for optimum adjustment of the above parameters are given; they also demonstrate the superiority of the improved EGPRS protocol stack versus the original GPRS one. Additionally, a protocol performance improvement both for GPRS and EGPRS has been proposed by introducing a new concept of "RLC frequent retransmission prevention" on the sender side, allowing fast polling by simultaneously avoiding retransmission of RLC blocks with pending acknowledgement status. Fast polling is referred to as the introduction of a very short RLC polling period, with, e.g., 8 RLC blocks, versus moderate polling periods, e.g., of 16 or 24 RLC blocks. The proposed strategy for fast polling proves to provide a user throughput gain in the order of 20% for GPRS and 30% for EDGE.