| Summary: | 碩士 === 國立臺灣科技大學 === 電機工程系 === 99 === The demand of design quality of service (QoS) mechanism for internetworking services is very imperious task. This motivation is interrelated with the problem of network bandwidth tendency to be very limited. For instance, multimedia application such as Video Streaming is necessarily to be transmitted in larger committed bandwidth quota, while ocassionaly undesired application flows prone to occupy most of available bandwidth and give some negative effects such as low-quality transmission impacting a low QoS for preferred application. Therefore providing isolation for bandwidth occupancy of each application is the solution that offered in this study. Exploiting Type of Service (ToS) field on the IPv4 header, each service can be identified its traffic behavior, then can be guaranteed by certain level of rate. The rate limiting is the strategy that purposed to obtain desired QoS of end to end application traffic flow.
Due to the sensitiveness of QoS metric, implementing rate limiting QoS design in hardware is considered to more accurate rather than merely perform simulation analysis. Although the commercial network devices such as router and switch are closed for any extension of QoS design and mechanism. Therefore the design is implemented usingOpenFlow System in NetFPGA platform. The modification of wildcard table module inOpenFlow user data path allow each packet is budgeted with a speed level on when it resides on the output queues.
After performing the test-bed implementation of service-based QoS mechanism, the system can isolate the traffic and limit the rate of traffic follow the determined bandwidth budget. The test-bed is also extended to wireless environment for considering QoS guarantee for mobile users. The performance result shows that the rate in the limitation of 10M, 1M and 10K, hits the average accurency of 85.6%, 96.2% and 78.3% for TCP (FTP) case and all hits the accurency of 100% for UDP (Video Stream) case; the percentage of packet loss is lower than without the mechanism 4.6% and the delay time is lower than without the mechanism 46.9%. As the overall the system offers better network utilization and QoS guarantee environment.
|
|---|
