Burst Switching Assembly Algorithm With The Business Flow Of Self-similarity

OBS networks combine electronic switching and optical Packet switching's strongpoint, is a balanceable choice. The edge router on OBS networks is a bridge between OBS and legacy networks, the traffic enter the OBS from legacy networks network by edge router. Packets are routed into assembly unites according to the QoS grade and egress address, then assemble them into bursts and create relevant control packets, at last we forward them separately into transparent optical channel or optoelectronic channel. By this way, we increase the switching granularity, and bursts can be transmitted entirely in optical domain. The traffic access into OBS network, mainly from packet switching network (include LAN, WAN, ATM etc.). This kind of traffic is long-range dependence/self-similarity. This character results in increase of packet lost rate and queuing delay of buffer system. We can look OBS network as two networks, one is transparent optical network for transmitting data burst; the other is optical and electronic hybrid control network for transmitting burst head packet. Actually control network is a packet switching network. So the self-similar traffics also bring negative effect to OBS network performance. We shape traffic to reduce its self-similarity which will reduce the negative effect brought by self-similarity,.The assembly mechanism of edge router affects the character of OBS networks traffic. If we could use assembly algorithm to reduce self-similarity of traffic accessed into OBS networks, which is self-similar, we can decrease packet lost rate of control networks, reduce queuing delay and improve OBS network performance to same extent. In this dissertation, we fully deeply discussed the architecture of OBS networks, analyzed the factors which affect the performance of OBS network, and then deeply discuss and validate the relationship between self-similarity of network traffic and assembly algorithm through theoretic and simulated method. At last, we conclude the rules of self-similarity of burst which change with assembly algorithm and characteristic of legacy traffic, and propose a new assembly algorithm.