Design of an Interactive Video-on-Demand System

Transcription

1 130 IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 5, NO. 1, MARCH 2003 Design of an Interactive Video-on-Demand System Yiu-Wing Leung, Senior Member, IEEE, and Tony K. C. Chan Abstract We design an interactive video-on-demand (VOD) system using both the client server paradigm and the broadcast delivery paradigm. Between the VOD warehouse and the customers, we adopt a client server paradigm to provide an interactive service. Within the VOD warehouse, we adopt a broadcast delivery paradigm to support many concurrent customers. In particular, we exploit the enormous bandwidth of optical fibers for broadcast delivery, so that the system can provide many video program and maintain a small access delay. In addition, we design and adopt an interleaved broadcast delivery scheme, so that every video stream only requires a small buffer size for temporary storage. A simple proxy is allocated to each ongoing customer, and it retrieves video from the optical channels and delivers the video to the customer through an information network. The proposed VOD system is suitable for large scale applications with many customers, and it has several desirable features: 1) it can be scaled up to serve more concurrent customers and provide more video programs, 2) it provides interactive operations, 3) it only requires point-to-point communication between the VOD warehouse and the customer and it does not involve any network control, 4) it has a small access delay, and 5) it requires a small buffer size for each video stream. Index Terms Broadcast delivery paradigm, client-server program, video-on-demand. I. INTRODUCTION AN INTERACTIVE video-on-demand (VOD) system provides an electronic video rental service to geographically distributed customers [1]. It retrieves video programs from its storage and delivers them to the customers through an information network. The customers can select and watch video programs at their convenient time and place, and they can interact with the programs via interactive operations such as pause, fastforward, and rewind. A VOD system has to serve multiple customers concurrently, and therefore it must have a large enough capacity to provide multiple video streams. Many designs have been proposed for this purpose and they can be classified into two categories: client server design and broadcasting design. A. Client Server Design The client server design adopts the client server paradigm. The system is composed of one or more servers [2], [3]. It maintains a dedicated video stream for each ongoing customer. When the customer performs an interactive operation, the system retrieves and delivers the corresponding video for him. This de- Manuscript received February 29, 2000; revised May 30, The associate editor coordinating the review of this manuscript and approving it for publication was Dr. Thomas R. Gardos. The authors are with the Department of Computer Science, Hong Kong Baptist University, Kowloon Tong, Hong Kong ( ywleung@ comp.hkbu.edu.hk; tkcchan@hkbu.edu.hk). Digital Object Identifier /TMM sign can provide an ideal interactive service to the customers, but it needs dedicated resources (such as I/O bandwidth) to maintain a video stream for each ongoing customer. For large scale applications with many customers, this design requires large amount of resources. A client server design can use a batching policy [4] [8] to serve more concurrent customers. The main idea is that the system waits for a time interval (called batch window) to collect a batch of requests for a video program. Then the system creates one video stream for this program and multicasts it to a batch of customers. In this manner, one video stream can serve multiple customers simultaneously. However, the customers have to wait before starting a VOD session (the waiting time is called access delay) and they cannot perform (or can only perform some constrained) interactive operations. Several batching policies have been proposed in the literature and they are as follows. Dan et al. [5] proposed that when the system can establish a new video stream, it selects the batch with the largest number of waiting customers and creates a video stream to serve all the customers in this batch. This batching policy can minimize the mean access delay, but some customers may experience long access delay. Dan et al. [6] proposed to choose a shorter batch window for the more popular video programs. They developed an analytical model and determined the window size for each video program. Almeroth et al. [7] proposed a batching policy that supports some constrained interactive operations by buffering a certain portion of the video program or joining the customer to another existing video stream. Liao and Li [8] proposed a batching policy called split-and-merge. When a customer performs an interactive operation, the system splits him from his original video stream and attempts to create a new video stream for him. If this is not possible, the customers waits. Once the interaction is done, the system attempts to merge this customer back to an existing video stream via buffering. If this is not possible, the customer waits. B. Broadcasting Design The broadcasting design adopts the broadcast delivery paradigm [9], [10] to serve many concurrent customers. There are three broadcasting designs for VOD [11] [13]. The first design is called periodic broadcasting [11]. It broadcasts multiple streams of the same video program at staggered times periodically. To watch a video program, a customer waits until a new video stream for this program is broadcast and then he receives this stream. The system can serve many concurrent customers because many customers can receive the same video stream from a broadcast channel simultaneously. However, it has a long mean access delay or /03$ IEEE

2 LEUNG AND CHAN: DESIGN OF AN INTERACTIVE VOD SYSTEM 131 Fig. 1. VOD system architecture. it requires many broadcast channels per video program. For example, if a 90-min video program is broadcast every 10 min, the mean access delay is 5 min and it requires nine broadcast channels. To support interactive operations, a low-resolution version of each video program is prepared and stored, and it is delivered to the customers upon requests using a client server paradigm. The second design is called staggered VOD [12]. It is similar to periodic broadcasting, but it provides interactive operations in a different manner. To perform an interactive operation, a customer changes to receive another broadcast video stream if it exists. To produce a good interactive effect, the staggering interval should be small but this would require a large number of broadcast channels. For example, if the staggering interval is 1 min, then a 90-min video program requires 90 broadcast channels. The third design is called pyramid broadcasting [13]. It divides each video program into segments of increasing sizes, and broadcasts the th segment of all the video programs periodically in the th broadcast channel. The bit rate of each broadcast channel is significantly larger than the video playback rate, especially when the number of video programs is large. For this reason, every customer must have a fast receiver and a large storage space. C. A New Design In this paper, we adopt both the client server paradigm and the broadcast delivery paradigm to design an interactive VOD system for large scale applications with many customers. This system has several desirable features. The system can be scaled up to serve more concurrent customers and provide more video programs. The system provides interactive operations which are approximations of the ideal ones. The customer can control the pause duration, the fast forward rate and the fast rewind rate. Fig. 2. Multiple optical fibers can be used to provide a large number of video programs. Every proxy is tapped to one of the optical fibers. Every proxy is tapped to any one of the optical fibers via a directional coupler. The system only involves point-to-point communication between the VOD warehouse and a customer. This type of communication can be supported by many existing network infrastructures. In contrast, some existing VOD designs assume that the network can support multicasting or broadcasting. The system does not involve any network control. This is important when the network is not owned and managed by the VOD service provider. In contrast, some existing designs involve network control such as dynamic multicasting. The access delay is small (say, 30 s). In contrast, the access delay in the client server designs with batching policies and the broadcasting designs is significantly longer (say, several minutes).

3 132 IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 5, NO. 1, MARCH 2003 Each video stream only requires a small buffer size for temporary storage. II. SYSTEM DESIGN A. Basic System Architecture Fig. 1 shows the proposed VOD system. The customers are connected to a VOD warehouse through an information network. The information network can be a private network owned by the VOD service provider, or it can be a public network. A customer makes requests via a low bit rate channel, and the VOD warehouse delivers the requested video to this customer via a high bit-rate channel. Both channels are point-to-point. Within the VOD warehouse, the video archives store video programs. Each video program is organized into pages, where every page lasts for the same duration. The video archives are connected to an optical fiber, which provides logical channels by wavelength division multiplexing [14]. The pages of each video program are read from the storage and are broadcast cyclically over multiple optical channels according to a broadcast delivery scheme. This scheme specifies the time and channel for broadcasting every page. We will design two broadcast delivery schemes in Section III. There are proxies tapped to the optical fiber, where a proxy is a simple logical unit for reception and transmission. When a customer initiates a new VOD session, the system allocates a free proxy to him. The proxy receives the requested video from the optical channels, and transmits it to the customer at the video playback rate through an information network. When the customer terminates the VOD session, the associated proxy will be released. B. Scalability As the VOD service becomes more popular, the system has to serve more concurrent customers and provide more video programs. The proposed VOD system is scalable to cope with these future expansions. To serve more concurrent customers, we add more proxies. It is not necessary to modify the existing hardware, and it is only necessary to modify the software setting to manage a larger number of proxies (e.g., to record which proxies are free). To provide more video programs, we add storage and optical fibers if the existing ones are not sufficient. When there are multiple fibers, each proxy can be tapped to one of them [Fig. 2], or it can be tapped to any one of them via a directional coupler [15] [Fig. 2]. If the VOD service covers a wide area (e.g., an entire country), we can replicate the VOD warehouse in distributed sites and each warehouse serves its nearest customers. This can avoid using long-distance channels, and reduce the propagation delay to give a better response time for interactive operations. III. BROADCAST DELIVERY SCHEMES In this section, we design two broadcast delivery schemes for broadcasting each video program over the optical channels. Fig. 3. Basic broadcast delivery scheme. These schemes are called basic broadcast delivery and interleaved broadcast delivery. We let be the bit rate of each optical channel and be the video playback rate. A. Basic Broadcast Delivery For clarity, we adopt the following case for explanation. A video program consists of pages and these pages are broadcast over channels. Generalization to any and is straightforward. 1) Delivery Schedule: Fig. 3 shows the basic broadcast delivery scheme. Time is divided into cycles where all the pages are broadcast once in a cycle. Each cycle is further divided into slots where one page is broadcast in one slot in a channel. We let the duration of a cycle and a slot be and respectively where. The first three pages are broadcast in channel 1 one after the other. Similarly, the next three pages are broadcast in channel 2; and the last three pages are broadcast in channel 3 (see Fig. 3). Fig. 4 shows how the proxy retrieves the pages from the optical channels, and Fig. 4 shows how to deliver these pages to the customer. First of all, the proxy tunes its receiver to channel 1, and waits until the beginning of the coming cycle. Then it retrieves page 1 from channel 1, and at the same time delivers this page to the customer at the video playback rate (a slower rate) through an information network. After retrieving page 1, the proxy does not retrieve in the next slots but it continues to deliver the remaining portion of page 1 to the customer [see Fig. 4]. In the second and third cycles, the proxy does similar steps to retrieve and deliver page 2 and page 3 respectively, and the details are shown in Fig. 4 and. After retrieving page 3 from channel 1, the proxy does not retrieve in the next slots (i.e., cycle 4) but it continues to deliver the remaining portion of page 3 to the customer. At the same time, it tunes its receiver to channel 2 so that it will be able to retrieve page 4, page 5, and page 6 in the fifth, sixth, and seventh cycles respectively. Then, the proxy tunes its receiver to channel 3 to retrieve page 7, page 8, and page 9 in the ninth, tenth, and eleventh cycles respectively. In this manner, the proxy can deliver video to the customer continuously [Fig. 4]. 2) Buffer Size, Tuning Time, and Slot Duration: The proxy retrieves video at the channel bit rate (say, 50 Mbps) for one slot and then waits for one cycle. At the same time, it continuously delivers video at the video playback rate (say, 1.5 Mbps).

4 LEUNG AND CHAN: DESIGN OF AN INTERACTIVE VOD SYSTEM 133 Fig. 4. Basic broadcast delivery for VOD. Proxy retrieves the shaded pages. Proxy delivers the retrieved pages to the customer. Therefore, it must have buffer for temporary storage. To determine the buffer size required, we analyze the buffer occupancy in the proxy. Fig. 4(c) shows the buffer occupancy versus time, as follows. In the first slot of the first cycle, the proxy retrieves page 1 and delivers it to the customer simultaneously. Since the retrieval rate is faster than the delivery rate, the buffer occupancy is increasing with time until it reaches a maximum at the end of this slot. The maximum buffer occupancy is equal to retrieval rate delivery rate duration of a slot. In the subsequent slots, the proxy does not retrieve any page but it still delivers the remaining portion

5 134 IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 5, NO. 1, MARCH 2003 (c) Fig. 4. (Continued.) Basic broadcast delivery for VOD. (c) Buffer occupancy in the proxy versus time. The buffer size required by the proxy is R T. of page 1 to the customer. Therefore, the buffer occupancy is decreasing with time until the end of the first slot of the second cycle. At this time instance, the buffer occupancy is zero. The above pattern is repeated and the details are shown in Fig. 4(c). Since the maximum buffer occupancy is, the proxy requires a buffer size of bits. When the proxy has retrieved all the pages from one channel, it tunes its receiver to another channel. We see from Fig. 4 that tuning must be done within one cycle. Therefore, the maximum permissible tuning time is seconds. is at least several seconds in practice (see the design examples in Section V-C), while the tuning time of the current optical receivers ranges from several milliseconds to several microseconds [15]. Therefore, the tuning speed is not a concern in practice. The slot duration depends on, and. It can be expressed in terms of these quantities as follows. A cycle has p slots and its duration is, and so the slot duration is.a proxy retrieves a page from an optical channel in one slot at rate, and so a page contains bits. The proxy delivers this page to the customer in slots at rate, and so a page contains bits. By equating these two quantities, we have. Therefore, the duration of a slot is equal to. B. Interleaved Broadcast Delivery In the previous studies [11], [16], interleaving was proposed to reduce the buffer size for VOD storage systems. In this sub- Fig. 5. Interleaved broadcast delivery scheme. section, we apply interleaving to enhance the basic broadcast delivery scheme so that the buffer size required by each proxy can be reduced. For clarity, we adopt the following case for explanation. A video program consists of pages and these pages are broadcast over channels. Generalization to any and is straightforward. 1) Delivery Schedule: The main idea is to divide each page into minipages, and interleave them in a cycle so that each proxy is only required to store at most one minipage at any time. Consequently, the proxy only requires a small buffer size. Page is divided into minipages, which are referred to as minipages. Time is divided into cycles, a cycle is divided into slots, and a slot is further divided into minislots where a minipage is broadcast in one minislot in a channel. Fig. 5 shows the interleaved broadcast delivery scheme. In every cycle, three consecutive pages are broadcast in each

6 LEUNG AND CHAN: DESIGN OF AN INTERACTIVE VOD SYSTEM 135 Fig. 6. Interleaved broadcast delivery for VOD. Proxy retrieves the shaded pages. channel, and their minipages are interleaved over nine minislots. Specifically, we broadcast the minipages of pages 1, 2, and 3 in channel 1, as follows: broadcast their first minipages one after the other (i.e., minipage, then, and then ); broadcast their second minipages one after the other (i.e., minipage, then, and then ; broadcast their third minipages one after the other (i.e., minipage, then, and then ). In channel 2, we broadcast the minipages of pages 4, 5, and 6 in a similar fashion but in a different order: broadcast their third minipages, then their first minipages, and then their second minipages. In channel 3, we broadcast the minipages of pages 7, 8, and 9 in another order: broadcast their second minipages, then their third minipages, and then their first minipages. Fig. 6 shows how the proxy retrieves the minipages from the optical channels. In the first cycle, it retrieves minipage in the first minislot, then retrieves minipage in the fourth minislot, and then retrieves minipage in the seventh minislot. In the second and third cycles, the proxy retrieves minipages,,,,, in a similar manner, and the details are shown in Fig. 6. After retrieving all the minipages from channel 1, the proxy tunes its receiver to channel 2. It retrieves minipage in the fourth minislot of the fourth cycle, then retrieves minipage in the seventh minislot of the fourth cycle, and then retrieves minipage in the first minislot of the fifth cycle. The proxy retrieves minipages,,,,, in a similar fashion, and the details are shown in Fig. 6. After retrieving all the minipages from channel 2, the proxy tunes its receiver to channel 3 to retrieve minipages. The details are shown in Fig. 6. We note that the proxy waits for minislots after retrieving the last minipage of a page, and it waits for minislots after retrieving any other minipage. To ensure continuous video delivery, a page (or minipages) must last for one cycle and one minislot. The proxy delivers the minipages to the customer as follows. When it starts to retrieve a minipage, it starts to deliver this minipage to the customer. It continuously delivers the minipages

7 136 IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 5, NO. 1, MARCH 2003 TABLE I BASIC BROADCAST DELIVERY VERSUS INTERLEAVED BROADCAST DELIVERY (c) Fig. 6. (Continued.) Interleaved broadcast delivery for VOD. Proxy delivers the retrieved pages to the customer. (c) Buffer occupancy in the proxy versus time. The buffer size required by the proxy is equal to x = R T =m. to the customers one after the other, and Fig. 6 shows the details. 2) Buffer Size, Tuning Time, and Minislot Duration: To determine the buffer size required by each proxy, we analyze its buffer occupancy versus time as follows [see Fig. 6(c)]: In the first minislot, the proxy retrieves minipage and delivers it to the customer simultaneously. Since the retrieval rate is larger than the delivery rate, the buffer occupancy is increasing with time until the end of this minislot. At this time instance, the buffer occupancy is equal to retrieval rate delivery rate duration of a minislot. In the second and third minislots, the proxy stops retrieval but it is still delivering the remaining portion of minipage to the customer. Therefore, the buffer occupancy is decreasing with time until the end of the third minislot. At this time instance, the buffer occupancy is equal to delivery rate duration of two minislots. In the fourth minislot, the proxy retrieves and delivers minipage. As a result, the buffer occupancy is increasing with time again until the end of this minislot. At this time instance, the buffer occupancy is equal to retrieval rate delivery rate duration of a minislot. In the fifth and sixth minislots, the proxy stops retrieval but it is still delivering the remaining portion of minipage. Therefore, the buffer occupancy is decreasing with time until the end of the sixth minislot. At this time instance, the buffer occupancy is equal to delivery rate duration of two minislots. In the seventh minislot, the proxy retrieves and delivers minipage and hence the buffer occupancy is increasing with time again. At the end of this minislot, the buffer occupancy is equal to retrieval rate delivery rate duration of a minislot. In the subsequent three minislots, the proxy stops retrieval but continues delivery until all the bits in the buffer have been delivered. At the end of the first minislot of the second cycle, the buffer occupancy becomes zero. The above pattern of buffer occupancy repeats for the subsequent minipages [see Fig. 6(c)]. Since the maximum buffer occupancy is, the proxy requires a buffer size of bits. When the proxy has retrieved all the minipages from one channel, it tunes its receiver to another channel. We see from Fig. 6 that tuning must be done within minislots. In other words, the maximum permissible tuning time is seconds. The minislot duration depends on,, and.itcan be expressed in terms of these quantities as follows. A cycle has minislots and its duration is, and so the minislot duration is. A proxy retrieves minipages of a page from an optical channel in respective minislots at rate, and so these minipages contain a total of bits. The proxy delivers these minipages to the customer in minislots at rate, and so these minipages contain bits. By equating these two quantities, we have. Therefore, the duration of a minislot is equal to. C. Comparison We have presented two broadcast delivery schemes. These schemes require different buffer size per proxy and different permissible tuning time, and Table I shows a comparison. Interleaved broadcast delivery is more attractive as it requires a

8 LEUNG AND CHAN: DESIGN OF AN INTERACTIVE VOD SYSTEM 137 Fig. 7. Pause operation. Ideal pause operation. Approximate pause operation. Fig. 8. Fast forward operation. Ideal fast forward operation. Approximate fast forward operation. much smaller buffer size per proxy. In addition, as we will explain in Section IV, it can also support better interactive operations. IV. PROVISION OF INTERACTIVE OPERATIONS In this section, we describe a set of interactive operations which are approximations of the ideal ones, and explain how the proposed VOD system can provide these approximate operations. Our explanation will be based on the interleaved broadcast delivery scheme, but the ideas are also applicable to the basic broadcast delivery scheme. A. Pause Fig. 7 shows the ideal pause operation. When the customer issues a pause command, the playout point of the video program is frozen, and the pause duration can be any positive value. Fig. 7 shows the approximate pause operation, in which the pause duration is an integral multiple of the cycle duration.if is smaller, the approximate pause operation is more similar to the ideal one. The proposed VOD system can provide the approximate pause operation as follows. When a customer issues a pause command to the proxy at time, the proxy stops retrieval/delivery temporarily. When the customer issues a resume command at time, the proxy resumes retrieval/delivery at the playout point where it is paused. In other words, the proxy resumes retrieval/delivery at time where is an integer such that. B. Fast Forward Fig. 8 shows the ideal fast forward operation. When the customer issues a fast forward command, the video program is Fig. 9. Realization of approximate fast forward operation; the proxy only retrieves and delivers the shaded pages. Realization at the page level. Realization at the minipage level. played at a faster and constant rate. Fig. 8 shows the approximate fast forward operation. It plays a small portion of video at the normal rate, then skips a portion, then plays a small portion, and then skips a portion, etc. The approximate fast forward operation can be realized at the page level or minipage level. If it is realized at the page level, it plays a page and then skips some subsequent pages [e.g., see Fig. 9]; if it is realized at the minipage level, it plays a minipage and then skips some subsequent minipages [e.g., see Fig. 9]. In general, if the cycle duration is shorter, the approximation is closer to the ideal one. In particular, the realization at the minipage level is better than that at the page level because a minipage lasts for a shorter duration. C. Fast Rewind Fig. 10 shows the ideal fast rewind operation. When the customer issues a fast rewind command, the video program is played in the reverse order at a fast and constant rate. Fig. 10 shows the approximate fast rewind operation. It plays a small portion of video at the normal rate, then skips a previous portion, then plays a small portion, and then skips a previous portion, etc. The approximate fast rewind operation can be realized at the page level or minipage level. If it is realized at the page level, it plays a page and then skips some previous pages [e.g., see Fig. 11]; if it is realized at the minipage level, it plays a minipage and then skips some previous minipages [e.g., see Fig. 11]. In general, if the cycle duration is shorter, the approximation is closer to the ideal one. In particular, the realization at the minipage level is better than that at the page level because a minipage lasts for a shorter duration.

9 138 IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 5, NO. 1, MARCH 2003 Fig. 10. Fast rewind operation. Ideal fast rewind operation. Approximate fast rewind operation. Fig. 11. Realization of approximate fast rewind operation; the proxy only retrieves and delivers the shaded pages. Realization at the page level. Realization at the minipage level. V. DESIGN CONSIDERATIONS AND EXAMPLES A. Design Issues Proxy: Each proxy is a simple logical unit for reception and transmission. It receives one page per cycle and therefore its receiver is utilized in only one of the p slots in each cycle. Therefore, multiple proxies can share a receiver. They can also share a transmitter for outgoing delivery. For example, if ATM links at 155 Mbps/link are used for outgoing delivery and is 1.5 Mbps, 103 proxies can share a transmitter for one ATM link. To simplify implementation, multiple proxies can reside on the same physical unit for sharing. Cost-effectiveness: To serve one additional concurrent customer, it is only necessary to add one proxy and it is not necessary to add the other resources. The cost required is small. Therefore, the proposed system is more cost-effective for larger scale applications with many customers. Optical bandwidth: The proposed system uses an optical fiber within the VOD warehouse. An optical fiber is cheap, and its current price is about U.S. $0.2 per yard [17]. Nevertheless, its bandwidth is about GHz [17], and it can currently provide about one hundred OC-48 channels at 2.5 Gbps/channel [18], [19]. Even if each channel is operated at the OC-1 rate 50 Mbps/channel (OC-1 is mature and economical), an optical fiber can still provide a capacity of Mbps 5 Gbps. This is sufficient for the current application (e.g., see the design examples in Section V-C). I/O bandwidth: Given a broadcast delivery scheme, the pages or minipages can be stored in such a way that they are read sequentially from the storage for broadcasting. In this manner, the I/O bottleneck can be avoided. The I/O bandwidth required depends on the number of video programs, but it does not depend on the number of concurrent customers. Therefore, the proposed system is particularly suitable for large scale applications with many customers. 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 that the system can simply read from a disk for broadcasting over an optical channel. For example, if we use a disk with I/O speed 50 Mbps [20], we can operate each optical channel at the OC-1 rate 50 Mbps. In this manner, the system requires an array of small capacity disks. MPEG Video: MPEG is a common video compression standard [21]. It produces three types of frames: I frames, P frames, and B frames. A group of frames consists of an I frame and a certain combination of P frames and B frames, and each group is independent of the other groups. The proposed system can support MPEG video as follows. Within the VOD warehouse, each minipage contains an integral number of groups, and its size is larger than the average size of these groups by a specified percentage margin in order to accommodate their variable sizes. Between the VOD warehouse and the customer, a streaming protocol can be used to deliver the video to the customer [22]. Video playback rate and duration: In the proposed system, different video programs can occupy different number of optical channels. Therefore, it can accommodate video programs with different playback rate (e.g., 1.5 Mbps for MPEG-1 and 4 Mbps for MPEG-2) and different duration (e.g., 90 min and 120 min). B. Selection of Design Parameters 1) Cycle Duration : The determination of the cycle duration involves two conflicting factors, as follows. If is larger, a channel can broadcast more pages in a cycle and hence an optical fiber can broadcast more video programs. Therefore, a smaller number of optical fibers is needed. If is larger, the mean access delay is longer. It is because when a customer initiates a new VOD session, the associated proxy has to wait for an average of s before it can retrieve the first page of the requested video program. The service provider can specify an acceptable mean access delay. Then the cycle duration can be chosen to be. 2) Number of Optical Fibers : The number of optical fibers required depends on the number of video programs, the durations of these programs, and the cycle duration. Each optical channel can broadcast bits in a cycle. If the th

10 LEUNG AND CHAN: DESIGN OF AN INTERACTIVE VOD SYSTEM 139 video program has a duration of s, it has bits and hence it requires channels where TABLE II SPECIFICATION FOR EXAMPLE 1 (1) Each video program is broadcast in one optical fiber. When, an optical fiber can already broadcast all the video programs and hence. Otherwise, multiple optical fibers are needed. The problem of minimizing is equivalent to the bin packing problem [23] [25] and this problem is NP-hard. When the number of video programs is not large (say, ), the solution space is not very large and it is possible to find the minimal. For example, we can formulate the problem as an integer programming problem and then apply the branch-and-bound technique [26] to find the optimal solutions. When is large, we have to resort to a good heuristic algorithm. Many heuristic algorithms have been proposed for bin packing, and the most famous one is probably the first-fit algorithm [23], [24]. This algorithm has time complexity, and it works as follows. Examine the video programs one after the other. For each video program, select the optical fiber that can first accommodate it. If all the existing optical fibers cannot accommodate this video program, introduce a new optical fiber for it. There are many other good heuristic algorithms, and a recent one can be found in [25]. 3) Number of Proxies : If the system has more proxies, it can serve more ongoing customers and hence it is less likely that a request for a new VOD session is blocked. However, the system is more costly. To make a tradeoff, the service provider can specify an acceptable blocking probability and then determine the minimal number of proxies required. We assume that the arrivals of requests for new VOD sessions follow a Poisson process with rate [5], [6], and we let be the mean duration of a VOD session (it is different from the mean duration of the video programs because the customers may perform interactive operations). The system of proxies can be modeled as an queue [27]. When all the proxies are serving customers, any additional request will be blocked and the blocking probability can be found to be [27]: (e.g., contain at least one group of nine frames for MPEG [21]). To minimize the buffer size required by each proxy, we choose the largest such that these constraints are fulfilled. C. Design Examples We simulate and test the proposed system using OPNET. In this subsection, we present two design examples. Example 1: We design a VOD system for a large scale application, in which there is one customer requesting a new VOD session in every second on average (i.e., request/s ) and the mean duration of a VOD session is 100 minutes (i.e., min). We adopt the specification in Table II. In addition, video is compressed by MPEG with nine frames per group, and each minipage contains at least two groups of frames with 10% margin. We determine the design parameters as follows. Given s, we choose the cycle duration to be s Each video program requires channels. There are 50 video programs. Therefore, the system requires two optical fibers. To fulfill the given requirement, the number of proxies must satisfy To fulfill the given requirement, we determine the smallest such that. Since is a decreasing function of, we can use the bisection method [28] to determine the minimal. 4) Number of Minipages Per Page : In interleaved broadcast delivery, a page is divided into minipages. To reduce the buffer size required by each proxy, should be as large as possible (see Table I). However, has to fulfill two constraints. First, the actual tuning time must be equal to or smaller than the maximum permissible tuning time. Second, each minipage may have to contain at least a certain number of frames (2) The minimal satisfying the above inequality can be found to be Since the tuning time cannot be larger than (see Table I), we have or. Since each minipage contains at least two groups of frames, or. We choose. The buffer size required by each proxy is Kbytes. Example 2: This example is the same as Example 1, except that the acceptable mean access delay is 5 s. We choose the cycle duration to be s. Then each video program requires 18 optical channels. Therefore, the system requires ten optical fibers, where each optical fiber accommodates five video programs. This design adopts a smaller cycle duration

11 140 IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 5, NO. 1, MARCH 2003 than the previous design in Example 1. Consequently, it provides a better quality (i.e., shorter access delay and better interactive operations) at the expenses of using more optical fibers. VI. CONCLUSIONS In this paper, we adopted both the client server paradigm and the broadcast delivery paradigm to design a VOD system for large scale applications with many customers. This system has the following desirable features. The system can easily be scaled up to serve more concurrent customers and provide more video programs. The system can provide interactive operations which are approximations of the ideal ones. The customer can control the pause duration, the fast forward rate and the fast rewind rate. The system only involves point-to-point communication between the VOD warehouse and each customer. This type of communication can be supported by many existing network infrastructures. The system does not involve any network control. This is important when the network is not owned and managed by the VOD service provider. The access delay is small. Each video stream only requires a small buffer size for temporary storage. REFERENCES [1] D. Deloddere, W. Verbiest, and H. Verhille, Interactive video on demand, IEEE Commun. Mag, pp , May [2] R. Tewari, R. Mukherjee, and D. M. Dias, Real-time issue for clustered multimedia servers,, IBM Res. Rep. RC 20020, [3] D. N. Serpanos, L. Georgiadis, and T. Bouloutas, MMPacking: A load and storage balancing algorithm for distributed multimedia servers, IEEE Trans. Circuits Syst. Video Technol, vol. 8, pp , Feb [4] D. P. Anderson, Metascheduling for continuous media, ACM Trans. Comput. Syst., vol. 11, no. 3, pp , August [5] A. Dan, D. Sitaram, and P. Shahabuddin, Scheduling policies for an on-demand video server with batching, in Proc. ACM Multimedia Conf. Expo., Oct 1994, pp [6] A. Dan, P. Shahabuddin, D. Sitaram, and D. Towsley, Channel allocation under batching and CVR control in video-on-demand systems, J. Parallel Distrib. Comput., vol. 30, pp , [7] K. C. Almeroth and M. H. Ammar, The use of multicast delivery to provide a scalable and interactive video-on-demand service, IEEE J. Select Areas Commun., vol. 14, no. 6, pp , Aug [8] W. J. Liao and V. O. K. Li, The split-and-merge (SAM) protocol for interactive video-on-demand systems, in Proc. IEEE INFOCOM, Japan, Apr [9] J. W. Wong, Broadcast delivery, Proc. IEEE, vol. 76, no. 12, pp , Dec [10] T. F. Bowen, G. Gopal, G. Herman, T. Hickey, K. C. Lee, W. H. Mansfield, J. Raitz, and A. Weinrib, The datacycle architecture, Commun. ACM, vol. 35, no. 12, pp , [11] T. Chiueh, A periodic broadcasting architecture for large-scale residential video-on-demand service, Proc. SPIE, vol. 2615, pp , Oct [12] R. O. Banker et al., Method of providing video on demand with VCR like functions, U.S. Patent , [13] S. Viswanathan and T. Imielinski, Metropolitan area video-on-demand service using pyramid broadcasting, Multimedia Syst., vol. 4, pp , [14] C. A. Bracket, Dense wavelength division multiplexing networks: Principles and applications, IEEE J. Select. Areas Commun., vol. 8, no. 6, pp , Aug [15] M. S. Borella, J. P. Jue, D. Banerjee, B. Ramamurthy, and B. Mukherjee, Optical components for WDM lightwave networks, Proc. IEEE, vol. 85, no. 8, pp , August [16] B. Ozden, R. Rastogi, and A. Silberschatz, On the design of a low-cost video-on-demand storage system, Multimedia Syst., vol. 4, pp , [17] S. Chatterjee and S. Pawlowski, All-optical networks, Commun. ACM, vol. 42, no. 6, pp , June [18] P. E. Green, Optical networking update, IEEE J. Select. Areas Commun., vol. 14, no. 5, pp , June [19] G. Wilfong, B. Mikkelsen, C. Doerr, and M. Zirngibl, WDM cross-connect architectures with reduced complexity, IEEE J. Lightwave Technol., vol. 17, no. 10, pp , Oct [20] Seagate Barracuda Disk Drive Product Manual, [21] B. Furht, A survey of multimedia compression techniques and standards. Part II: Video compression, Real-Time Imag., vol. 1, pp , [22] M. Reisslein and K. W. Ross, High performance prefetching protocols for VBR prerecorded video, IEEE Network, vol. 12, no. 6, pp , Dec [23] R. K. Ahuja, T. L. Magnanti, and J. B. Orlin, Network Flows: Theory, Algorithms and Applications. Englewood Cliffs, NJ: Prentice-Hall, 1993, p. 87. [24] T. C. Hu, Combinatorial Algorithms. Boston, MA: Addison-Wesley, 1982, pp [25] A. Scholl, R. Klein, and C. Jurgens, BISON: A fast hybrid procedure for exactly solving the one-dimensional bin packing problem, Comput. Oper. Res., vol. 24, no. 7, pp , [26] G. L. Nemhauser and A. W. Lawrence, Integer and Combinatorial Optimization: Wiley, [27] D. Gross and C. M. Harris, Fundamentals of Queueing Theory, 2nd ed. New York: Wiley, 1985, pp [28] S. Yakowitz and F. Szidarovszky, An Introduction to Numerical Computations, 2nd ed. New York: MacMillan, Yiu-Wing Leung (M 92 SM 96) received the B.Sc. and Ph.D. degrees from the Chinese University of Hong Kong in 1989 and 1992, respectively. His Ph.D. advisor was Prof. T. S. Yum. After graduation, he was with the Hong Kong Polytechnic University until He then joined the Hong Kong Baptist University and now he is an Associate Professor in the Department of Computer Science. He has a wide range of research interests and he has been working on three main areas: information networks, multimedia communications and systems, and cybernetics and systems engineering. He has published more than 50 journal papers in these areas, most of which appear in various IEEE publications. He has supervised several Ph.D. students and postdoctoral fellows, and some of them now teach at various universities in Singapore, the United Kingdom, Canada, and Hong Kong. Tony K. C. Chan received the M.Sc. degree in information technology from the Department of Computing, Hong Kong Polytechnic University. He is currently pursuing the Ph.D. degree in the Department of Computer Science at Hong Kong Baptist University. His professional interests include information networks and multimedia systems.