Comprehensive statistical admission control for streaming media servers

Abstract

Streaming media servers and digital continuous media recorders require the scheduling of I/O requests to disk drives in real time. There are two accepted paradigms to achieve this: deterministic or statistical. The deterministic approach must assume larger bounds on such disk parameters as the seek time, the rotational latency and the transfer rate, to guarantee the timely service of I/O requests. The statistical approach generally allows higher utilization of resources, in exchange for a residual probability of missed I/O request deadlines. We propose a novel statistical admission control algorithm called TRAC based on a comprehensive three random variable (3RV) model to support both reading and writing of multiple variable bit rate media streams on current generation disk drives. Its major distinctions from previous work include (1) a very realistic disk model which considers multi-zoning of disks, seek and rotational latency profiles, and unequal reading and writing data rate limits, (2) a dynamic bandwidth sharing mechanism between reading and writing, and (3) support for random placement of data blocks. We evaluate the TRAC algorithm through an extensive numerical analysis and real device measurements. The results show that it achieves a much more realistic resource utilization (up to 38% higher) as compared with the best, previously proposed algorithm based on a single random variable (1RV) model. Most impressive, in all the experiments the difference between the results generated by TRAC and the actual disk device measurements match closely.

Adaptive Threshold-Based Admission Control for Distributed Video-on-Demand Systems

“…Many admission control algorithms have been proposed and can be classified into deterministic [1] [2], observation-based [3] and statistical ones [4] [5]. But all of these papers focus on improving the QoS assurance and do not consider the inherent property of client requests such as the priority of clients.…”