Scheduling and bandwidth allocation for the distribution of archived video in VOD systems

Providing costeffective videoondemand (VOD) services necessitates reducing the required bandwidth for transporting video over highspeed networks. In this paper, we investigate efficient schemes for transporting archived MPEGcoded video over a VOD distribution network. A video stream is characterized by a timevarying traffic envelope, which provides an upper bound on the bit rate. Using such envelopes, we show that video streams can be scheduled for transmission over the network such that the perstream allocated bandwidth is significantly less than the source peak rate. In a previous work [13], we investigated stream scheduling and bandwidth allocation using global traffic envelopes and homogeneous streams. In this paper, we generalize the scheduling scheme in [13] to include the heterogeneous case. We then investigate the allocation problem under windowbased traffic envelopes, which provide tight bounds on the bit rate. Using such envelopes, we introduce three streamscheduling schemes for multiplexing video connections at a server. The performance of these schemes is evaluated under static and dynamic scenarios. Our results indicate a significant reduction in the perstream allocated bandwidth when stream scheduling is used. While this reduction is obtained through statistical multiplexing, the transported streams are guaranteed stringent, deterministic quality of service (i.e., zero loss rate and small, bounded delay). In contrast to video smoothing, our approach requires virtually no buffer at the settop box since frames are delivered at their playback rate.     

Building video‐on‐demand servers

We have designed and implemented a controllable software architecture for a VideoonDemand (VOD) server. With the proposed software architecture, many system design issues can be investigated. For example, we studied several disk striping schemes in the storage subsystem and examined the impact of the disk striping schemes on the utilization of system resources. In the processing component, we observed that additional concurrent video streams can be supported by using efficient memory interleaving. Buffering with a large memory space in the processing subsystem is a common technique to alleviate the latency variance of accessing different system components. By employing userlevel control and scheduling, the variance can be decreased thereby reducing the resulting buffer space needed for each video stream. In the network subsystem, we adopted a serverdriven approach for investigating MPEG2 video delivery over Asynchronous Transfer Mode (ATM) networks. The VOD server controls the pace of video transmission and reduces the complexity of the client. Since the client has limited buffer space (due to cost considerations), we have reduced the buffer requirement by regulating the transmission based on timing information embedded in the MPEG2 streams. Our research and experimental results are based on a VOD server which is currently under construction. The prototype server is based on an SGI sharedmemory multiprocessor with a mass storage system consisting of RAID3 disk arrays. Using 30 RAID3 disk arrays, preliminary experimental results show that the prototype server can potentially support more than 360 highquality video streams with careful design and coordination of the different system components.     

Dynamic capacity allocation and hybrid multiplexing techniques for ATM wireless LANs

We consider digital wireless multimedia LANs and timevarying traffic rates. To deal effectively with the dynamics of the timevarying traffic rates, a Traffic Monitoring Algorithm (TMA) is deployed to dynamically allocate channel capacities to the heterogeneous traffics. The TMA is implemented as a higher level protocol that dictates the capacity boundaries within two distinct framed transmission techniques: a Framed Time DomainBased (FTDB) technique and a Framed CDMA (FCDMA) technique. The performance of the TMA in the presence of the FTDB technique is compared to its performance in the presence of the FCDMA technique for some traffic scenarios. The performance metrics used for the TMAFTDB and TMAFCDMA combinations are channel capacity utilization factors, traffic rejection rates, and traffic delays. It is found that the TMAFTDB is superior to the TMAFCDMA when the speed of the transmission links is relatively low and the lengths of the transmitted messages are relatively short. As the speed of the transmission links and the length of the transmitted messages increase, the TMAFCDMA eventually outperforms the TMAFTDB.     

NETMAN: The design of a collaborative wearable computer system

This paper presents a wearable groupware system designed to enhance the communication and cooperation of highly mobile network technicians. It provides technicians in the field with the capabilities for realtime audioconferencing, transmission of video images back to the office, and contextsensitive access to a shared notebook. An infrared locationtracking device allows for the automatic retrieval of notebook entries depending on the user's current location.     

A Markov modulated fluid queueing system with strict priority

We investigate a Markov modulated fluid queueing system with strict priority. The input process is composed of two fluid flows which are stored in buffer1 and buffer2, respectively. The rates of these fluid flows depend on the current state of a finite state Markov chain. Buffer1 has full assignment of priority (=strict priority) for service and so buffer2 is served at a residual service rate when buffer1 is empty. We explicitly derive the stationary joint distribution of the two buffer contents in the system by a spectral decomposition method. In the case of a twostate Markov chain, the joint distribution is explicitly expressed in terms of the system parameters. Also the joint moments and tail distributions of the two buffer contents are obtained and some numerical examples are presented.     

Tree multicast strategies in mobile,multihop wireless networks

Tree multicast is a well established concept in wired networks. Two versions, persource tree multicast (e.g., DVMRP) and shared tree multicast (e.g., Core Based Tree), account for the majority of the wireline implementations. In this paper, we extend the tree multicast concept to wireless, mobile, multihop networks for applications ranging from ad hoc networking to disaster recovery and battlefield. The main challenge in wireless, mobile networks is the rapidly changing environment. We address this issue in our design by: (a) using soft state (b) assigning different roles to nodes depending on their mobility (2level mobility model); (c) proposing an adaptive scheme which combines shared tree and persource tree benefits, and (d) dynamically relocating the shared tree Rendezvous Point (RP). A detailed wireless simulation model is used to evaluate various multicast schemes. The results show that persource trees perform better in heavy loads because of the more efficient traffic distribution; while shared trees are more robust to mobility and are more scalable to large network sizes. The adaptive tree multicast scheme, a hybrid between shared tree and persource tree, combines the advantages of both and performs consistently well across all load and mobility scenarios. The main contributions of this study are: the use of a 2level mobility model to improve the stability of the shared tree, the development of a hybrid, adaptive persource and shared tree scheme, and the dynamic relocation of the RP in the shared tree.     

Simulation‐based performance evaluation of routing protocols for mobile ad hoc networks

In this paper we evaluate several routing protocols for mobile, wireless, ad hoc networks via packetlevel simulations. The ad hoc networks are multihop wireless networks with dynamically changing network connectivity owing to mobility. The protocol suite includes several routing protocols specifically designed for ad hoc routing, as well as more traditional protocols, such as link state and distance vector, used for dynamic networks. Performance is evaluated with respect to fraction of packets delivered, endtoend delay, and routing load for a given traffic and mobility model. Both small (30 nodes) and medium sized (60 nodes) networks are used. It is observed that the new generation of ondemand routing protocols use much lower routing load, especially with small number of peertopeer conversations. However, the traditional link state and distance vector protocols provide, in general, better packet delivery and endtoend delay performance.     

The effects of asymmetry on TCP performance

In this paper, we study the effects of network asymmetry on endtoend TCP performance and suggest techniques to improve it. The networks investigated in this study include a wireless cable modem network and a packet radio network, both of which can form an important part of a mobile ad hoc network. In recent literature (e.g., [18]), asymmetry has been considered in terms of a mismatch in bandwidths in the two directions of a data transfer. We generalize this notion of bandwidth asymmetry to other aspects of asymmetry, such as latency and mediaaccess, and packet error rate, which are common in widearea wireless networks. Using a combination of experiments on real networks and simulation, we analyze TCP performance in such networks where the throughput achieved is not solely a function of the link and traffic characteristics in the direction of data transfer (the forward direction), but depends significantly on the reverse direction as well. We focus on bandwidth and latency asymmetries, and propose and evaluate several techniques to improve endtoend performance. These include techniques to decrease the rate of acknowledgments on the constrained reverse channel (ack congestion control and ack filtering), techniques to reduce source burstiness when acknowledgments are infrequent (TCP sender adaptation), and algorithms at the reverse bottleneck router to schedule data and acks differently from FIFO (acksfirst scheduling).     

A general purpose proxy filtering mechanism applied to the mobile environment

Host mobility complicates the standard networking model in unexpected ways. It increases network heterogeneity, causing difficulty within applications that expect a high or constant amount of network bandwidth. Mobility increases cost when a user switches from a relatively inexpensive high bandwidth network to a low bandwidth paybycall or paybybyte network. Mobility may also reduce security when a user moves from a safe network onto a foreign one. This situation can be improved by introducing a proxy between the mobile client and its peer. The purpose of this intermediary is to process data moving between the client and peer. The type of processing performed depends on the desired result: the proxy can hide network heterogeneity via data filtering, reduce cost by using compression, and increase security through the use of secure protocols between the client and proxy. This paper presents a general purpose proxy system and how it has been applied to the mobile environment. The overall architecture is described, examples of its use given, and a study on its feasibility and performance is presented.     

Floor acquisition multiple access (FAMA) in single‐channel wireless networks

The FAMANCS protocol is introduced for wireless LANs and adhoc networks that are based on a single channel and asynchronous transmissions (i.e., no time slotting). FAMANCS (for floor acquisition multiple access with nonpersistent carrier sensing) guarantees that a single sender is able to send data packets free of collisions to a given receiver at any given time. FAMANCS is based on a threeway handshake between sender and receiver in which the sender uses nonpersistent carrier sensing to transmit a requesttosend (RTS) and the receiver sends a cleartosend (CTS) that lasts much longer than the RTS to serve as a busy tone that forces all hidden nodes to back off long enough to allow a collisionfree data packet to arrive at the receiver. It is shown that carrier sensing is needed to support collisionfree transmissions in the presence of hidden terminals when nodes transmit RTSs asynchronously. The throughput of FAMANCS is analyzed for singlechannel networks with and without hidden terminals; the analysis shows that FAMANCS performs better than ALOHA, CSMA, and all prior proposals based on collision avoidance dialogues (e.g., MACA, MACAW, and IEEE 802.11 DFWMAC) in the presence of hidden terminals. Simulation experiments are used to confirm the analytical results.     

MA‐WATM: A new approach towards an adaptive wireless ATM network

In a cellular multimedia network like wireless ATM (WATM), self control seems primordial. Our new approach is based on the application of DAI (distributed artificial intelligence) techniques in order to build a selfadaptive network within random nonuniform traffic conditions. Attempting to achieve a high network capacity in terms of resource allocation and air interface BER (bit error rate), we propose to apply intelligent agent features to enhance the architecture of WATM systems. In fact, the intelligent platform MAWATM (multiagent wireless ATM) is used to provide a flexible integration of the multiagent technique in wireless entities, BSs (base stations), MSs (mobile stations) and MSCs (mobile switching centers). Two applications are investigated and explored by simulations in order to bear out the feasibility of the multiagent approach.     

Comparative analysis of wireless ATM channel access protocols supporting multimedia traffic

Extension of multimedia services and applications offered by ATM networks to wireless and mobile users has captured a lot of recent research attention. Research prototyping of wireless ATM networks is currently underway at many leading research and academic institutions. Various architectures have been proposed depending on the intended application domain. Successful implementation of wireless connectivity to ATM services is significantly dependent on the medium access control (MAC) protocol, which has to provide support for multimedia traffic and for qualityofservice (QoS) guarantees. The objective of this paper is to investigate the comparative performance of a set of access protocols, proposed earlier in the literature, with more realistic source traffic models. Data traffic is modeled with selfsimilar (fractal) behavior. Voice traffic is modeled by a slow speech activity detector (SAD). Video traffic is modeled as a H.261 video teleconference, where the number of ATM cells per video frame is described by a gamma distribution and a firstorder discrete autoregressive process model. A comparison of the protocols based on simulation data is presented. The goal of the paper is to identify appropriate techniques for effectively and efficiently supporting multimedia traffic and QoS. Simulation results show that boundaries between different types of services are necessary for multimedia traffic. Reservation for certain traffic type especially video can significantly improve its quality. Reducing the number of collisions is an important issue for wireless networks since contentions lead not only to potentially high delay but also result in high power consumption.     

Evaluating scheduling in integrated‐services networks using a semi‐fluid‐flow model

An integratedservices network carries diverse traffic, which leads to diverse performance objectives. For example, voice and video packets typically have performance objectives based on the fraction of packets that will be delivered within a given delay bound, while data packets often have objectives based on mean delay. Greater loads can be supported in networks in which a voice or video packet is given priority over data packets if and only if the former is in danger of missing its deadline. Algorithms that allow this include CostBased Scheduling, occupancybased algorithms, the Priority Token Bank, and to a lesser extent, the Leaky Bucket. This paper presents an approach to evaluating performance with these algorithms in the realistic case where data arrivals are highly bursty, but voice and video packets are not. Mean queueing delay for data bursts is determined analytically in some important scenarios, and an efficient simulation approach based on the same model is described for cases where analysis is not currently possible. The model is a semifluidflow model in which voice or video packets are assumed to arrive as a continuous fluid flow, whereas data packets arrive in large bursts at discrete instants in time.     

A scalable wireless virtual LAN

This paper presents a Wireless Virtual Local Area Network (WVLAN) to support mobility in IPoverATM local area networks. Mobility is handled by a joint ATMlayer handoff for connection rerouting and MAClayer handoff for location tracking, such that the effects of mobility are localized and transparent to the higherlayer protocols. Different functions, such as Address Resolution Protocol (ARP), mobile location, and ATM connection admission are combined to reduce protocol overhead and frontend delay for connectionless packet transmission in connectionoriented ATM networks. The proposed WVLAN, through the use of ATM technology, provides a scalable wireless virtual LAN solution for IP mobile hosts.     

Personal Imaging and lookpainting as tools for personal documentary and investigative photojournalism

A means and apparatus for covert capture of extremely highresolution photorealistic images is presented. The apparatus embodies a new form of userinterface – instead of the traditional point and click metaphor which was thought to be the simplest photography had to offer, what is proposed is a look metaphor in which images are generated through the natural process of looking around, in a manner that does not require conscious thought or effort. These lookpaintings become photographic/videographic memories that may, at times, exceed the quality attainable with even large and cumbersome professional photographic film cameras, yet they may be captured through a device that resembles ordinary sunglasses. The method is based on longterm psychophysical adaptation using a covert sunglassbased realitymediating apparatus, together with two new results in image processing. The first new result is a means of estimating the true projective coordinate transformation between successive pairs of images, and the second is that of estimating, to within a single unknown scalar constant, the quantity of light arriving at the image plane. Furthermore, what is captured is more than just a picture. The resulting environment map may be explored by one or more remote participants who may also correspond and interact with the wearer during the actual shooting process, giving rise to computer supported collaborative (collective) photography, videography, shared photographic/videographic memory, etc.     

Demand re‐registration for PCS database restoration

In a Personal Communications Services (PCS) network, mobility databases such as Home Location Register (HLR) and Visitor Location Register (VLR) are utilized to support mobility management for Mobile Stations (MSs). If the location databases fail, the location information loss or corruption will seriously degrade the service offered to the subscribers. In this paper, we propose a new VLR failure recovery scheme called demand reregistration. In this scheme, the VLR broadcasts a reregistration request to all MSs after the VLR failure. When an MS receives the reregistration request, it sends a registration message to the VLR to recover the location record. Since all MSs will reregister after receiving the broadcasting request, traffic jam (and thus collisions) may occur. If a collision occurs, then the involved MSs must resend the registration messages. This paper studies the performance of demand reregistration by investigating how effectively the reregistration can recover the location record for an MS before the first MS call termination occurs. Our results indicate that demand reregistration can effectively recover VLR failure.     

An adaptive algorithm for optimizing the packet size used in wireless ARQ protocols

We develop an algorithm that allows an ARQ protocol to dynamically optimize the packet size based on estimates of the channel biterrorrate. Our algorithm is particularly useful for wireless and satellite channels where the biterrorrates tend to be relatively high and time variable. Our algorithm uses the acknowledgment history to make estimates of the channel biterrorrate, based on which the optimal packet size can be chosen. We develop a Markov chain model for the analysis of the system, under static channel conditions, and show that the algorithm can achieve close to optimal performance using a history of just 10,000 bits. We also use the Gilbert–Elliott twostate Markov channel to model dynamic channel conditions. We show, through simulation, that the algorithm performs well even under rapidly changing channel conditions. Finally, we discuss a maximum likelihood approach for choosing the packet size, which performs almost optimally but is much easier to implement.     

Throughput performance of distributed applications in switched networks in presence of end‐system bottlenecks

The steady state throughput performance of distributed applications deployed in switched networks in presence of endsystem bottlenecks is studied in this paper. The effect of various limitations at an endsystem is modelled as an equivalent transmission capacity limitation. A class of distributed applications is characterised by a static traffic distribution matrix that determines the communication between various components of the application. It is found that uniqueness of steady state throughputs depends only on the traffic distribution matrix and that some applications (e.g., broadcast applications) can yield nonunique values for the steady state component throughputs. For a given switch capacity, with traffic distribution that yield fair unique throughputs, the tradeoff between the endsystem capacity and the number of application components is brought out. With a proposed distributed rate control, it has been illustrated that it is possible to have unique solution for certain traffic distributions which is otherwise impossible. Also, by proper selection of rate control parameters, various throughput performance objectives can be realised.     

Consistency management in Deno

We describe a new replicatedobject protocol designed for use in mobile and weaklyconnected environments. The protocol differs from previous protocols in combining epidemic information propagation with voting, and in using fixed perobject currencies for voting. The advantage of epidemic protocols is that data movement only requires pairwise communication. Hence, there is no need for a majority quorum to be available and simultaneously connected at any single time. The protocols increase availability by using voting, rather than primarycopy or primarycommit schemes. Finally, the use of perobject currencies allows voting to take place in an entirely decentralized fashion, without any server having complete knowledge of group membership. We show that currency allocation can be used to implement diverse policies. For example, uniform currency distributions emulate traditional voting schemes, while allocating all currency to a single server emulates a primarycopy scheme. We present simulation results showing both schemes, as well as the performance advantages of using currency proxies to temporarily reallocate currency during planned disconnections. Furthermore, we discuss an initial design of the underlying replicatedobject system and present a basic API.     

Time‐varying modeling of a nonstationary signal

In this paper, a method for modeling a nonstationary signal using timevarying parameters by considering that the signal is generated by a linear, timevarying (LTV) system with a stationary white noise input is presented. This method is based on the Wold–Cramer (WC) representation of a nonstationary signal. Because the relationship between the generalized transfer function of an LTV system and the timevarying coefficients of the difference equation of a discretetime system is not addressed so far in the literature, therefore, in this paper a solution to this problem is proposed. A simple relationship between the system generalized transfer function and the timevarying parameters of the system is derived, then an MLS algorithm is developed to solve for the system timevarying parameters. Computer simulation illustrating the effectiveness of our algorithm is presented.