Design of an Interactive Video- on-Demand System Yiu-Wing Leung, Senior Member, IEEE, and Tony K. C. Chan IEEE Transactions on multimedia March 2003

Outline Introduction VOD system architecture Broadcast delivery schemes Interactive operations Design considerations and examples Conclusions

Introduction Client-Server Design Maintain a dedicated video stream for each customer Use batching policy to server more concurrent customers Customers must wait before starting a VOD session (called access delay)

Introduction Broadcasting Design Periodic broadcasting Broadcast multiple streams of the same video at staggered times periodically Staggered broadcasting Similar to periodic, but perform an interactive operation

VOD system architecture Video archives Connect to an optical fiber and provide logical channels Contain a lot of videos Broadcast over multiple optical channels according to a broadcast delivery scheme Proxy Logical unit for reception and transmission Receives the video from optical channel, and transmits it with video playback rate

VOD system architecture Scalability To add storage and optical fibers if not sufficient VOD warehouse in distributed site and nearest customers

Broadcast delivery schemes Each video is organized into pages A video consists of n=9 pages and these are broadcast over C=3 channels Two types of broadcast delivery schemes Basic broadcast delivery Interleaved broadcast delivery

Basic broadcast delivery Video archives broadcast diagram

Basic broadcast delivery Proxy receives the shaded pages

Basic broadcast delivery Proxy delivers the retrieved pages to the customer

Basic broadcast delivery Buffer size Proxy retrieves video at the channel bit rate (50Mbps), and delivers video at the video playback rate (1.5Mbps), so must have temporary storage Maximal buffer size (R c - R v ) * (Tc / p) = R v * Tc Retrieval rate : R c Delivery rate : R v Duration of a slot : Tc / p

Basic broadcast delivery Tuning time When proxy has retrieved all the pages from one channel, it tunes its receiver to another channel. The maximum permissible tuning time is Tc seconds. Slot duration Depend on Tc, R c, R v Proxy retrieves a page from channel in one slot : (R c Tc / p) bits Proxy delivers this page to the customer in p+1 slots : R v (p+1)Tc/p bits => Tc / p = Tc*R v / (R c – R v )

Interleaved broadcast delivery Divide each page into m minipages, and interleave them in a cycle. Page i divided into m minipages, referred to as minipages i 1, i 2, …,i m

Interleaved broadcast delivery A page (or m minipages) must last for one cycle and one minislot

Interleaved broadcast delivery Proxy delivers the retrieved pages to the customer

Interleaved broadcast delivery Buffer size (1) x1 = (R c – R v ) * (Tc / mp) = R v * Tc( 1+1/mp-1/p) / m (2) y1 = (x1 – R v * (2Tc / mp) ) = R v * Tc / m 2 p (3) x2 = (y1 + (R c – R v ) * (Tc / mp) ) = R v * Tc( 1+2/mp-1/p) / m (4) y2 = (x2 – R v * (2Tc / mp) ) = 2R v * Tc / m 2 p (5) X3 = (y2 + (R c – R v ) * (Tc / mp) ) = R v Tc / m Maximum buffer size is R v Tc / m

Interleaved broadcast delivery Tune time Proxy retrieved all the minipages of a page from one channel. Tuning must be done within p minislots. Maximum permissible tuning time is (Tc / mp) * p = Tc / m Minislot duration Depends on Tc, R c, R v, m Proxy retrieves m minipages of a page from an optical channel : R c Tc / p bits Proxy delievers m minipages to the customer in mp+1 minislots : R v *(mp+1)Tc / mp => Tc / mp = Tc R v / (m*R c – R v )

Comparision Interleaved broadcast scheme support better interactive operations

Interactive operations Pause : Tc is smaller, the approximate is more similar to the ideal one Fast forward : (1) Play a small portion of video at normal rate (2) Minipage level is better than page level

Interactive operation Fast rewind :

Design consideration Design issue Optical bandwidth : an optical fiber provide 5 Gbps, so if one channel needs 50 Mbps, and can provide 100 channel to use. I/O speed and channel bit rate : we can match the I/O speed of a disk with the bit rate of an optical channel, so system requires an small capacity disk. Video playback rate and duration : Different video can occupy different number of channels, therefore can accommodate video with different playback rate (e.g., MPEG-1 and MPEG 2) and different duration (e.g., 90min and 120min).

Design consideration Design parameters Cycle duration Tc If Tc is larger, a channel can broadcast more pages in a cycle If Tc is larger, the mean access delay is longer. If service can specify an acceptable mean access delay T *, then Tc can be chosen to 2T * Number of minipages per page m A page divide into m minipages can reduce each proxy buffer size The actual tuning time must be equal to or smaller than the maximum permissible tuning time Tc / m Each minipage may have to contain at least a certain number of frames (e.g., contain at least one GOP of nine frame for MPEG)

Design example 1 Video is compressed by MPEG with nine frames per GOP. Because T * =30s, so Tc=2T * =60S Each video require [(1.5*106*90*60) / (50*106*60)] = 3 channels. There are 50 video program, so require two optical fibers Because tuning time cannot not be larger than Tc / m, so 10*10 -3 ≤ 10/m => m ≤ 1000 But since each minipage contain at least two GOP of frames Tc ≥ m( 2*9 / 30 ) => m ≤ 100 Buffer size : R v Tc / m = Kbytes

Design example 2 Change acceptable mean access delay T * =5s, so Tc=2*5=10 Each video program requires 18 optical channels, so requires ten optical fibers, where each optical fiber accommodates five video programs. Consequently, it provides a better quality (i.e., shorter access delay and better interactive operation, but use more optical fibers.

Conclusion Adopt both the client-server paradigm and the broadcast delivery paradigm. The system can easily be scaled up to serve more concurrent customer and provide more video. Provide interactive operations which are approximations of the ideal ones. The access delay is small Each video stream only requires a small buffer size for temporary storage.