A large‐scale multiple‐access method for video‐on‐demand systems

Abstract

SUMMARYA multiple-access method is proposed for simultaneously providing video data to more than 100 terminals in a video-on-demand system. The proposed method divides up each video program into segments, stores the segments on multiple disk drives, and perform time-division-multiplex reading based on the time-slot sequences for each disk drive that are periodically phase shifted with respect to each other. This method features: 1) theoretical limit on the maximum number of multiple accesses that can be implemented, and straightforward upscaling of the maximum number of multiple accesses because the amount of buffer memory needed for each terminal is fixed; 2) no request loss, even when many terminals request the same program because the load for reading each video program is always distributed across all the disk drives; and 3) system control that guarantees read-request response times within a specified value by allowing the skipping of a very small amount of leading information in each video program. Numerical examples show that this access method makes it possible to design a system having no request losses and a guaranteed response time within one second, even for simultaneous accesses by 900 terminals.

Evaluation of disk‐retry methods for continuous media data

“…P effect for DCP. = 0, 3,10,256). This is because with DCP, since only the retries are executed in cases when time cannot be taken for parity calculations, when the error rate increases, there will no longer be opportunities in which parity calculations are executed and no differences will appear.…”

“…The top part of Fig. 5 shows the assignment of a timeslot to each client [10]. As the figure shows, each client is assigned an interval as a timeslot until the deadline is reached, and data is read within that interval.…”

Traffic characteristics of striped disks in video servers with visual search functions

“…2,3 This suggests that if programs are stored on individual disks using a conventional computer file system, the workload will be heavily concentrated on the disks storing the popular shows, which would severely limit the number of simultaneous requests that could be supported by the system. 4 To overcome this problem, striping schemes have been proposed that divide a movie up into small video segments which are then evenly distributed ͑or striped͒ for storage across multiple disks. [4][5][6][7][8] Because video materials are striped across the component disks without regard to the frequency with which the programs are accessed, this always results in an even distribution of accesses across the disks in the server.…”

“…4 To overcome this problem, striping schemes have been proposed that divide a movie up into small video segments which are then evenly distributed ͑or striped͒ for storage across multiple disks. [4][5][6][7][8] Because video materials are striped across the component disks without regard to the frequency with which the programs are accessed, this always results in an even distribution of accesses across the disks in the server. So ͑theoretically at least͒ it should be possible to expand the number of accesses without limit simply by adding additional disks to the server.…”

“…So ͑theoretically at least͒ it should be possible to expand the number of accesses without limit simply by adding additional disks to the server. 4 This is true if terminals are limited to normal playback capability; however, if jump, skip search, and other visual search functions are implemented, then accesses could be briefly concentrated on a particular disk, which could cause interruptions in playback at terminals. 5,6,8 It is thus extremely important that the potential adverse impact of this sort of momentary workload skewing be taken into consideration in the design of video servers based on striped disk arrays.…”