Recovery Performance of PGM with Redundancy Estimation for Proactive FEC
Sonia Fahmy, Renin M. Jegadeesan · Purdue e-Pubs (Purdue University System) · 2003
We study the recovery performance of the Pragmatic General Multicast (PGM) reliable multicast protocol using adaptive proactive forward error corrcclion (FEe). We compare the performance of PGM wilbout FEe, PGM with static proactive FEe, PGM with previously proposed redundancy estimation algorithms, and PGM with II new redundancy estimation technique which we call smooth increase/decrease (smooth-ID). Our technique periodically estimates whether the network is becoming more or less congested/crror-prone based on loss rates, and determines the amount of redundancy information to include based upon this estimate. Temporary fluctuations of the loss rates do not cause drastic increases or dccrcascs of the amount of redundancy. We evaluate the performance wilh different group memberships, different loss models and congested links, and different transmission group (TG) sizes. Our results show that PGM with our redundancy estimation technique performs best in terms of high recovery percentage and low overhead. We also find that the PGM transmit window size and advancc frcquency are kcy parameters, and affect the choice of transmission group size. KeywordsForward Error Correction (FEC), proactive FEC, Pragmatic General Mullicast (pGM), reliable multicasf, multicasting, error recovery