Jayanta K. Dey

Online smoothing of variable-bit-rate streaming video

Bandwidth smoothing techniques for stored video perform end to end workahead transmission of frames into the client playback buffer, in advance of their display times. Such techniques are very effective in reducing the burstiness of the bandwidth requirements for transmitting compressed, stored video. This paper addresses online bandwidth smoothing for a growing number of streaming video applications such as newscasts, sportscasts, and distance learning, where many clients may be willing to tolerate a playback delay of a few seconds in exchange for a smaller bandwidth requirement. The smoothing can be performed at either the source of the videocast or at special smoothing server(s) (e.g., proxies or gateways) within the network. In contrast to previous work on stored video, the online smoothing server has limited knowledge of frame sizes and access to only a segment of the video at a time. This is either because the feed is live or because it is streaming past the server. We formulate an online smoothing model which incorporates playback delay, client and server buffer sizes, server processing capacity, and frame size prediction techniques. Our model can accommodate an arbitrary arrival process. Using techniques for smoothing stored video at the source as a starting point, we develop an online, window-based smoothing algorithm for delay tolerant applications. Extensive experiments with MPEG-1 and M-JPEG video traces demonstrate that online smoothing significantly reduces the peak rate, coefficient of variation, and effective bandwidth of variable-bit-rate video streams. These reductions can be achieved with modest playback delays of a few seconds to a few tens of seconds and moderate client buffer sizes, and closely approximate the performance of optimal offline smoothing of stored video. In addition, we show that frame size prediction can offer further reduction in resource requirements, though prediction becomes relatively less important for longer playback delays. However, the ability to predict future frame sizes affects the appropriate division of buffer space between the server and client sites. Our experiments show that the optimal buffer allocation shifts to placing more memory at the server as the server has progressively less information about future frame sizes.

Online smoothing of live, variable-bit-rate video

Bandwidth smoothing techniques are effective at reducing the burstiness of a compressed, pre-recorded video stream by prefetching frames into the client playback buffer in advance of each burst. In contrast to stored video, live applications typically have limited knowledge of frame sizes and often require bounds on the delay between the source and the client(s). This paper addresses bandwidth smoothing for a growing number of live video applications, such as videocasts of courses or television news, where many clients be willing to tolerate a playback delay of several seconds or minutes in exchange for a smaller throughput requirement. Extending techniques for smoothing pre-recorded video, we develop online, window-based smoothing algorithms for these delay-tolerant applications. Experiments with MPEG traces demonstrate that the new algorithms significantly reduce the peak rate, coefficient of variation, and effective bandwidth of variable-bit-rate video streams using fairly small window sizes (1-10 seconds), closely approximating the performance of the optimal offline algorithm.

Playback restart in interactive streaming video applications

Low latency is crucial in networked multimedia applications such as on-demand streaming audio and video. In this paper we consider the problem of restarting or resuming playback following an interactive operation such as fast forward, rewind, or indexed jump in stored on-demand video. We present two approaches to restart playback after an interactive operation, and develop algorithms to compute the latencies incurred in restarting playback. Using long MPEG-1 traces, with an optimal smoothing technique to transmit data for regular playback, we demonstrate that the latencies incurred under our approaches are very small. We next examine the latencies incurred when restarting playback in a video server under two policies for sharing its bandwidth: (a) one in which only regular playback bandwidth is available to restart playback and (b) one in which additional free server bandwidth (if any) is accessible. We find playback restart latencies to be similarly low under both policies, suggesting that the simpler approach (a) is sufficient.

Streaming CBR transmission of VBR stored video

Data transmission across a network using constant-bit-rate (CBR) service simplifies admission control and resource management techniques. We consider lossless, starvation- free, streaming CBR transmission of compressed digital video, which is known to exhibit significant, multi-time- scale rate variability. This transmission uses work-ahead transfer into available client buffers to send data at a rate significantly below the peak rate of the original video. The goal of any video transmission scheme is to minimize resources requirements such as client buffer, transmission rate, channel holding time and playback startup latency. We identify, for CBR video transmissions, formal structural properties of the tradeoffs among these resources. Specifically, we show that, (i) the minimum feasible client buffer requirement as a function of playback startup latency is unimodal with one minimal value, (ii) the minimum feasible CBR rate is a convex decreasing function of the startup latency, and (iii) the corresponding channel holding time is piecewise linear concave increasing function of the startup latency. Using these structural properties, we then develop an O(N log N) algorithm that computes the minimum client buffer size and the associated CBR rate and playback startup latency required to transmit a VBR video. This is a significant improvement over an existing O(N2 log N) algorithm to solve the same problem. We next quantitatively examine the resource tradeoffs using MPEG-1 traces, and find that both the CBR transmission rate and minimum client buffering requirement can be substantially reduced by requiring only very small playback startup latencies.

Optimal multicast smoothing of streaming video over an internetwork

A number of applications such as internet video broadcasts, corporate telecasts, distance learning etc. require transmission of streaming video to multiple simultaneous users across an internetwork. The high bandwidth requirements coupled with the multi-timescale burstiness of compressed video make it a challenging problem to provision network resources for transmitting streaming multimedia. For such applications to become affordable and ubiquitous, it is necessary to develop scalable techniques which can efficiently deliver streaming video to multiple heterogeneous clients across a heterogeneous internetwork. We propose using multicasting of smoothed video and differential caching of the video at intermediate nodes in the distribution tree, as techniques for reducing the network bandwidth requirements of such dissemination. We formulate the multicast smoothing problem, and develop an algorithm for computing the set of optimally smoothed transmission schedules for the tree (such that the transmission schedule along each link in the tree has the lowest peak rate and rate variability for any feasible transmission schedule for that link) given a buffer allocation to the different nodes in the tree. We also develop an algorithm to compute the minimum total buffer allocation to the entire tree and the corresponding allocation to each node, such that feasible transmission is possible to all the clients, when the tree has heterogeneous rate constraints. MPEG-2 trace-driven performance evaluations indicate that there are substantial benefits from multicast smoothing and differential caching. For example, the optimal multicast smoothing can reduce the total transmission bandwidth requirements in the distribution tree by more than a factor of 3 as compared to multicasting the unsmoothed stream.

Storage subsystem in a large multimedia server for high-speed network environments

Multimedia information systems are an essential component of several multimedia applications ranging from media entertainment to on-line library systems. However, the bandwidth available on traditional workstation-based I/O systems is not sufficient to be able to support hundreds of multimedia information streams. In this paper, we discuss our approach to built a storage subsystems in a multimedia server which can deliver variable-bit-rate audio/video data streams to several hundred clients simultaneously over a high-speed broadband network. Our approach is to distribute the bandwidth requirement across multiple disks in a disk-array architecture. We address issues of multimedia data layout on this disk subsystem.

Optimal multicast smoothing of streaming video over the Internet

A set of applications such as Internet video broadcasts, corporate telecasts, and distance learning require the simultaneous streaming of video to a large population of viewers across the Internet. The high bandwidth requirements and the multi-timescale burstiness of compressed video make it a challenging problem to provision network resources for streaming multimedia. For such applications to become affordable and ubiquitous, it is necessary to develop scalable techniques to efficiently stream video to a large number of disparate clients across a heterogeneous Internet. In this paper, we propose to multicast smoothed video over an application-level overlay network of proxies, and to differentially cache the video at the intermediate nodes (proxies) in the distribution tree, in order to reduce the network bandwidth requirements of video dissemination. We formulate the multicast smoothing problem as an optimization problem, and develop an algorithm for computing the set of transmission schedules for the tree that minimize the peak rate and rate variability, given buffer constraints at different nodes in the tree. We also develop an algorithm to compute the minimum buffer allocation in the entire tree, such that feasible transmission to all the clients is possible, when the tree has heterogeneous rate constraints. We show through trace-driven simulations that substantial benefits are possible from multicast smoothing and differential caching, and that these gains can be realized even with modest proxy caches.