| Multimedia services such as Video-on-Demand (VoD) are gaining in popularity. Video data is time-constrained, bursty and requires massive storage space. Multiprocessors make good VoD servers since they can handle a large number of concurrent requests. Traditional multiprocessors use general-purpose network service algorithms that are not conducive for VoD. Simple fairness algorithms such as Round-Robin, provide no service guarantees and cannot manage the congestion caused by burstiness of data. Algorithms that provide network service guarantees such as Stop-and-Go, cannot manage the size of video data. Stop-and-Go's admission control policy leads to under-utilization of the network, poor throughput and excess total latency for video data.; We introduce a new algorithm called Frame-based Arbitration and Scheduling Technique (FAST), which is designed for handling video traffic. We show that FAST's policy of allowing bursty data into the network leads to superior throughput and total latency. The low queuing delay and packet loss figures also emphasize that FAST is suitable for use in multiprocessor-based VoD servers. |
