Publications

List of Publications

Research

Video-on-Demand

    Video on demand (VOD) is a technology for many important applications. A VOD system allows distributed clients to play back any video from a large collection stored on one or more servers.  To accept a user request, the VOD server must allocate enough resources to guarantee a jitter-free playback of the video.  Such resources include storage and network I/O bandwidth.  Sufficient storage bandwidth must be available for continuous transfer of data from storage to the network interface card (NIC), which in turn needs enough bandwidth to forward the stream to remote clients. To support a large number of users, two types of multicast have been studied.  In Non-periodic multicast environment, users make requests of videos to the server; and it serves requests according to some scheduling policy.  To conserve server bandwidth, requests made by several clients for the same video within a short period of time can be served as a group. In Periodic Broadcast environment, users do not make requests to the server.  Rather, the server broadcasts the video periodically.  Although, this type of technique does not guarantee true VOD, the worst service latency experienced by any client is always less than threshold.  A distinct advantage of this approach is that it can serve a very large community of users using minimal server bandwidth.  In fact, the bandwidth requirement is independent of the number of subscribers to the system. 

----To TOP---

VCR-like interactivity in a Periodic Broadcast Framework

In video-on-demand (VOD) applications, it is desirable to provide the user with the video-cassette-recorder-like (VCR) capabilities such as fast-forwarding a video or jumping to a specific frame.  In the broadcast framework,each video is broadcast repeatedly on the network.  Existing techniques rely on data prefetching as the mechanism to provide this functionality.  This approach provides limited usability since the prefetching rate cannot keep up with typical fast-forward speeds.  Fast-forwarding a video for several seconds would inevitably exhaust the prefetch buffer.  Because of the nature of the periodic broadcast paradigm, the server can periodically broadcast interactive versions of videos. For instance, an interactive version might contain only every fifth frame in the original video.  The client software can therefore leverages these “interactive” broadcasts to provide better VCR services. Our work is to investigate this new approach and compare it to data prefetching algorithms.

----To TOP---

Bandwidth Heterogeneity in Periodic Broadcast Video Streaming.

Essentially all existing periodic broadcast techniques require the VOD service providers to tailor their system design to fit the capability of a specific class of users.  This approach has two disadvantages.  First, users with lower bandwidth will not be able to watch the video as continuity cannot be guaranteed; and second users having more bandwidth at their disposal will not be able to benefit from their surplus.
To address the above drawbacks, we can consider encoding the video stream into a decomposition of layers.  Depending on the bandwidth available at the end user, more or less layers are received, resulting in a variation in the display quality.  This strategy can be adapted for periodic broadcast to provide adaptability to different receiving bandwidths.  This simple solution, however, would present some drawbacks for a high quality VOD system.  Firstly, a user with lesser bandwidth is enforced to sacrifice video quality; secondly, it is very demanding on user bandwidth to obtain a satisfying image quality; and finally, it is difficult to implement. One solution to this particular problem is to have a single broadcast strategy and different reception strategies tailored to the bandwidth capacity of several classes of clients. 

----To TOP---

Database Terrain Updates in Distributed Interactive Simulation.

In real-time interactive training and simulation systems, changes of the natural terrain surface as well as other non-moving objects are important to simulators. In fact, entities moving to new regions need to know the latest updates to the dynamic terrain. Two scalable schemes to communicate this information to moving entities are discussed. In the first scheme, changes of states of the dynamic terrain are multicast to entities, while the second sheme communicate the changes using a streaming periodic broadcast technique. Both techniques present good performance under different situations. Under heavy entity traffic the streaming periodic broadcast technique outperforms the multicast scheme, and if the for lower traffic the multicast-based technique outperforms periodic broadcast.

----To TOP---

P2P video streaming.

In Overlay Networks, the infrastructure-based approach lessens the bottleneck burden at the server side due to the fact that clients can get services not only from the server, but also from overlay nodes. However P2P video streaming presents important design issues. Firstly, P2P streaming systems should offer a short access latency (quick joining time), allowing new peers to receive the desired video quickly. Secondly, A quick and graceful recovery procedure is needed to handle peer failures. The failure recovery procedure should not only reconnect a disconnected peer to another peer, but must also quickly localize the failure such that only few peers are affected. Finally, Overhead for exchanging information among peers must be kept small. Existing techniques in the P2P approach can be categorized into techniques supporting live video streaming and those that support pre-recorded video streaming. Some techniques can offer both services. Live video streaming differs from pre-recorded video streaming in two important aspects. First, the access latency is more crucial to live video streaming than to pre-recorded video streaming. Second, a user joining a current streaming session of live-streaming is only concerned about a stream starting from his/her joining time. Third, degradation of video quality for live video streaming is crucial since the option of watching the video for a second time may not be available.
 
----To TOP---