Hop-by-hop flow control with packet aggregation in TCP/IP networks

Abstract

The internet is once again suffering from its own success. Since 1995, there has been a massive increase in demand for internet services, resulting in an exponential growth of the Internet. We have entered now an era where the users of the Internet are unable to obtain the bandwidth needed to support their applications, and they are experiencing high packet loss. Packet loss problem arises whenever the number of packets arriving at a given router is much higher than its buffering space. The main goal of this thesis is to find adaptive schemes capable of optimizing the communication network performance of systems with buffering constraints such as TCP/IP networks. A new strategy, called hop-by-hop flow-control with packet aggregation (HFCPA), is devised for optimizing the network performance. HFCPA is a variation of the current TCP/IP protocol. Flow control with aggregation of packets is implemented in the main routers of the network, in contrary to the current TCP/IP protocol where the flow control occurs at the edge of the network (i.e. at the end-users). In fact, the internet will be subdivided into tunnels where the edge routers of each tunnels are responsible for the packet management inside the tunnel (aggregation, fragmentation, bandwidth allocation, path selection...). With this newly proposed technique, we have shown that TCP connections experience lower packet loss and higher throughput compared with normal TCP/IP implementation, especially during network congestion.