The Application Of Flow Label On Quality Of Service Over Heterogeneous Networks

Future telecommunication networks are essentially composed of different portions and technologies: each single portion may be defined by different group and implement a different QoS solution. QoS requests should traverse the overall network from the source to the destination through portions that implement different technologies and different protocols. Therefore, QoS request should be received and understood by each specific portion where QoS may have a different meaning and interpretation. However, it is still an open question how to map different QoS class and parameters among different types of networks.After analyzing the current QoS mechnisms, flow identification is supposed to be a sticking point to solve the QoS problem over heterogeneous networks. In IPv6, two fields directly in the IP header have been defined to mark traffic flow, i.e., Flow Label field (20 bit) and Traffic Class field (8 bit), which can be used to support QoS management. The 20-bit Flow Label in addition to the 8-bit Traffic Class field can provide a huge traffic classification space. Flow Label locates in the main header of IP packet so that it can avoid encapsulation (as in MPLS) and fragmentation (as in ATM) of IP packets. Moreoever, no additional layers are required to add identification labels. Since the packet size does not contain additional labels, flow label's application eliminates the potential risk of MTU violations. It also allows node-local generation of labels, so it increases network flexibility and may also considerably enhance security.This thesis mainly studies the application of Flow label on QoS over heterogeneous networks, and has proposed practical schemes. Based on existing QoS module and Flow Label's features, two QoS modules are proposed in this thesis, which are negotiation-based QoS model and mapping-based QoS model. In the former model, QoS class interpretation can be conducted in per flow granularity. End-to-end QoS can be guaranteed through the negotiation and mapping of Flow Label. The latter deals with QoS class mapping in an opposite way, which can overcome the shortcomings of existing QoS Class mapping mechanisms. In the latter model, each QoS class does not need to find its corresponding class in all different networks anymore; rather, it only needs to find its position in a uniform QoS space: Flow Label. Each model has their strength points and drawbacks.In this thesis, the proposed solutions are evaluated by using OPNET. Some new network modules in OPNET are designed and implemented to support the flow label's application. Experiment results show that our proposed flow label based model is able to guarantee end-to-end QoS performance for QoS-specific traffic with per-flow identification ability and achieve better performance than pure DiffServ networks.