The Study on Quality of Service Guarantees of Compressed Video Sources in Packet Switch Networks

• Main Authors: Ren-Jing HUANG, 黃仁景
• Other Authors: Yen-Ping CHU
• Format: Others
• Language: zh-TW
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/60776818818481202588

博士 === 國立中興大學 === 應用數學系 === 90 === In this dissertation, we want to provide the packet switch network with an end-to-end delay quality of service (QoS) guarantee and utilize the bandwidth of the network more efficiently. Consider the network’s use of the FCFS switch nodes that provide the same delay for all the connections, it means the end-to-end delay of all the connections will be same if they go through the same switch nodes. The admission control functions in a network require an adequate per-switch delay allocation policy to allocate end-to-end delay requirements to a set of local switches. In this dissertation, we consider a deterministic service that provides a deterministic number of connections for each service class. We compare two kinds of allocation policies. One is EQ policy (Equal Allocation) where each switch node shares the time delay equally. The other is an OPT policy (Optimal Allocation) where each node can share a different time delay per its residential bandwidth. Therefore, it can avoid any switch node that uses all its allocated bandwidth such that no new connection can be enabled. In order to evaluate the performance of difference allocation policies, we define two performance indices; the admissible region (AR) and the relative gain (RG). In order to realize the packet flow in the above network, we wrote a program to simulate each packet flow in the regulators, delay jitters and FCFS switch nodes. With this simulation program, we evaluate the accumulated bits in the buffer of FCFS node and also the end-to-end delay of each packet. This information motivates the study of the second part of the dissertation. In the second part, we apply the above N service class networks to the MPEG video sources. We use a set of 10 minute long star wars MPEG-compressed video traces to evaluate our proposed networks. The networks use the leaky bucket to define the input traffic. Since the traffic of the MPEG packet burst rate varied with each of the 10 different transmissions, this presented us with the following problem. If we use the peak rate to determine bandwidth allocation, it will occupy too large network bandwidth and preclude new connections. If we use the mean rate to determine bandwidth allocation, it will introduce longer time delays and not meet the delay requirement. Therefore, we derived a closed form formula to determine the best rate to transfer the MPEG video and allow new connections equal to the network’s use of multiple leaky buckets.