Fair bandwidth distribution marking and scheduling algorithm in network traffic classification

The traffic classification has the ability to solve and manage many network difficulties and problems for network management and service providers. Despite these efforts, various challenging tasks still remain unsolved in DiffServ. The concept DiffServ network defines two main routers. First, an edge router that is involved in marking, shaping or issuing the policy for flows based on their SLAs. Second, core routers which offer the schedule of different traffic using the marks they carry from the edge router. However, DiffServ fails to provide individual flow guarantees, but it provides statistical guarantees of service classes. This thesis focuses on the fair share of excess bandwidth distribution among aggregates flow; First, analyse the effect of aggregate based marking schemes on the fair sharing of excess bandwidth among aggregates. Second, propose an Optimized Time Sliding Window based Three Colour Marker. Finally, propose a new method of obtaining optimal parameters dropping functions for Random Early Detection (RED) algorithm. The analytical model was improved, computing the marking probability can be used in the planning of a network architecture. They can be useful for taking a decision on choosing concrete values of traffic classification environments element parameters in a real network. Appropriate selection of configuration parameters of network elements is crucial for providing a certain level of QoS for customers as well as for fulfilling the SLA. Careful scrutiny of markers and droppers characteristics, as well as characteristics of a compound network architecture, supports such a decision. Therefore, in order to improve bandwidth fairness along with efficient optimization to alleviate this problem, we propose two marker algorithms; a Double Modified Double Improved time sliding window Three Colour Marker (M2I2TSWTCM) algorithm, which makes a new value of that depends on the logarithm peak information rate (PIR) and which is added to the adaptive factor that exists in the previous algorithm. Second, an Optimized time sliding window packet marker (OTSWTCM) algorithm. This algorithm depends on the adaptability of the concept in the ITSWTCM, I2TSWTCM and M2I2TSWTCM algorithms for affecting the fairness and multiple protocols. This is achieved by using a new design algorithm by separating marker TCP and UDP protocols and extended the marking probability for injecting more green and yellow traffic into the network. In addition, the marking probability mechanism was studied to check how the parameters in the traffic rate affect fairness. Additionally, the traffic are relying on the markers and scheduling algorithms to the service classes at the routers. The higher level priority agreements give a higher or equal probability than the lower level, this technique is perfect at a core router by scheduling algorithm. Thus, proposing the method of reestimating the dropping functions in the RED algorithm. This method manages resources at the core routers to guarantee the QoS. The main goal in this part is not only to improve bandwidth fairness, but also to improve average throughput and reduce the average delay and packet loss. More importantly, integrating the marker with scheduler algorithms in order to reduce the traffic loss and improve bandwidth fairness. Several simulation experiments and analytical models have been presented with respect to traffic classification environments and scenarios. This enables evaluating the performance of the proposed mechanisms compared with the existing various approaches. The simulation results show that the proposed mechanisms significantly improve edge and core routers’ performance at traffic classification network. The improvement goal is not only to fairly share excess bandwidth distribution among aggregates, but also to reduce the average delay and packet loss. Moreover, it satisfies the stringent requirements for supporting real-time services in terms of traffic loss, end-to-end delay as well as fairness.

Text FSKTM 2019 5 - ir.pdf Download (1MB)

Actions (login required)

View Item