On the Decoding Cost of Streaming Forward Erasure Correction Codes

Streaming forward erasure correction (FEC) is effective for achieving low-latency packet delivery over loss links. Two decoding strategies, namely the forward-first or backward-first Gaussian eliminations, are commonly used, which correspond to first constructing an upper or lower triangular decoding matrix, respectively. This letter is devoted to analyzing the decoding costs of the strategies. We show that the costs can be effectively analyzed using a queue theory modeling, and reveal that the backward-first and forward-first algorithms are more advantageous at lower and higher repair insertion percentages, respectively. The boundary where the two algorithms are approximately equal can also be identified.