Issue No.12 - December (2005 vol.16)
Zhenhai Duan , IEEE
Zhi-Li Zhang , IEEE
Yiwei Thomas Hou , IEEE
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/TPDS.2005.149
<p><b>Abstract</b>—In this paper, we investigate the fundamental trade-offs in aggregate packet scheduling for support of guaranteed delay service. In our study, we consider two classes of aggregate packet scheduling algorithms: the <it>static earliest time first</it> (SETF) and <it>dynamic earliest time first</it> (DETF). Through these two classes of aggregate packet scheduling (and together with the simple FIFO packet scheduling algorithm), we show that, with additional timestamp information encoded in the packet header for scheduling purposes, we can significantly increase the maximum allowable network utilization level, while, at the same time, reducing the worst-case edge-to-edge delay bound. Furthermore, we demonstrate how the number of the bits used to encode the timestamp information affects the trade-off between the maximum allowable network utilization level and the worst-case edge-to-edge delay bound. In addition, the more complex DETF algorithms have far superior performance than the simpler SETF algorithms. These results illustrate the fundamental trade-offs in aggregate packet scheduling algorithms and shed light on their provisioning power in support of guaranteed delay service.</p>
Packet scheduling algorithms, aggregate packet scheduling algorithms, quality of services (QoS), performance analysis.
Zhenhai Duan, Zhi-Li Zhang, Yiwei Thomas Hou, "Fundamental Trade-Offs in Aggregate Packet Scheduling", IEEE Transactions on Parallel & Distributed Systems, vol.16, no. 12, pp. 1166-1177, December 2005, doi:10.1109/TPDS.2005.149