A study of rate-based credit adaptation algorithm in the interworking system between IP network and ATM network

Abstract

The focus of this dissertation is the congestion management in the interworking system between IP network and ATM network. We suggest a new interworking model using proxy signaling agent(PSA) and rate-based credit adaptation(RBCA) algorithm, validating its algorithm through various simulations. When IP network and ATM network are interworked together, it is impossible to control the congestion and flow through end-to-end terminals. End user uses the TCP/IP protocol, so window-based congestion control algorithms(slow-start, congestion avoidance, fast retransmission, and fast recovery) are worked, while core network is the part of ATM that controls the congestion by mean of various rate-based algorithms. Therefore, crucial mismatches between rate-based control and window-based control take place. For example, there exists the edge router as an access point from IP network to ATM network, and rate-based controlled ATM switches maintain their buffer usage at the moderate level, pushing the traffic mostly to the edge router. Consequently, periodic overflow and underflow in the buffer of edge router takes place, and link utilization cannot become optimal level, at all. ATM protocol manages the in-service traffic by mean of UPC(Usage Parameter Control), that is, violation traffic of any prior traffic contract is filtered, smoothed, and at last, discarded. This imposes signal to the TCP to retransmit the discarded packet, and this situation occurs again and again. We resolve this problem by receiver’s window modification at the edge router. The edge router, named interworking unit(IWU), can be interoperable from IP to ATM by mean of PSA. Proxy signaling is the method that non-ATM user can use ATM fully without any signaling overhead. One of the advantages of this method is that no need to modify the legacy protocols and network facilities. Moreover, it has very simple procedure in comparison with the conventional, overlay, and peer model. Our new interworking model usi...