A Neural Network Based Proactive Buffer Control Approach for Better Reliability and Performance for Object-based Internet Applications

摘要

In this paper we propose a neural controller for adaptive buffer management. The aim is to maintain a safety marginA between the buffer length and the queue length so that buffer overflow at the reception side is eliminated. This elimination would lead to better reliability and performance for object-based applications over the Internet. Different preliminary test results confirm that the neural controller is indeed an effective solution and is more responsive than the previous P+I+D approach.