Practical Recovery Solution for Information Loss in Real-Time Network Environment

Abstract

Feedback mechanism based algorithms are frequently used to solve network optimization problems. These schemes involve users and network exchanging information (e.g. requests for bandwidth allocation and pricing) to achieve convergence towards an optimal solution. However, in the implementation, these algorithms do not guarantee that messages will be delivered to the destination when network congestion occurs. This in turn often results in packet drops, which may cause information loss, and this condition may lead to algorithm failing to converge. To prevent this failure, we propose least square (LS) estimation algorithm to recover the missing information when packets are dropped from the network. The simulation results involving several scenarios demonstrate that LS estimation can provide the convergence for feedback mechanism based algorithm.