Deep Reinforcement Learning-based Congestion Control Algorithms (DRL-based CCA) have shown their great potential to adapt to various environments automatically (e.g., Orca). However, it is difficult for existing DRL-based CCAs to achieve superior QoS consistently, particularly over mobile networks with rapid network fluctuations. The fundamental problem stems from the training of a single model that encompasses a wide range of network conditions. The resulting model can be overly generalized, rendering it less accurately or optimally tailored for a specific network condition. To tackle this challenge, we develop Network-segmented Model Specialization (NMS), a framework that automatically maximizes QoS for any DRLbased CCA under different network conditions. Specifically, NMS generates a set of models offline, each trained for a specific network segment. Online, it selects the model based on the current segment. We showed that NMS not only improves the QoSs of existing DRL-based CCAs consistently, but also opens a new way for the exploration of a clean-slate approach, as opposed to following the hybrid TCP/DRL approach. Hence, we designed Galaxy, a novel clean-slate DRL-based CCA that addresses the inherent limitations in clean-slate approaches by incorporating network segments' knowledge. Extensive evaluations show NMSoptimized Galaxy further explores NMS to achieve superior QoS.

Short-video services have seen explosive growth in recent years. Streaming over mobile networks is inherently challenging due to the latter's bandwidth fluctuations, motivating researchers to develop many sophisticated adaptive bitrate (ABR) algorithms to compensate. While ABR, together with prefetching, has been proposed for playlist streaming, its application to non-playlist streaming has received little attention. This work fills this gap by first exploring the efficacy of directly applying ABR to non-playlist streaming. Observing their limitations motivates the development of a new class of inter-stream bitrate adaptation (ISA) algorithms. Unlike ABR, ISA adapts bitrate on a per-video basis, which is not only simpler to implement and deploy but can even outperform ABR algorithms by up to 66.71% across a wide range of networks. Moreover, ISA and ABR are complementary such that they can be combined into Integrated Bitrate Adaptation (IBA) algorithms to raise performance gains further by up to 77.03%. In addition, this work develops a novel adaptive rebuffering duration (ARD) algorithm specifically designed for frame-based playback common in short-video services to further improve their performance under challenging network conditions. Together, ISA and ARD offer a new set of tools with progressive complexity-performance tradeoffs for enhancing the performance of short-video services.

Short video applications such as TikTok, Douyin, and Kwai have experienced significant popularity in recent years. However, the quality of experience (QoE) provided by short video streaming services still falls short of expectations. As a leading provider of short video services with proprietary video players and content delivery network (CDN) capabilities, we are in a unique position to optimize the QoE of these services. In this study, we present our pilot investigation into congestion control performance optimization for short video services by leveraging collaboration between user-end video players and edge servers. Based on a comprehensive measurement study of network characteristics from production networks and incorporating feedback from video players, we developed an optimized congestion control algorithm called BBR-E2E. We deployed BBR-E2E in our production network and conducted a large-scale A/B testing across the country, involving trillions of video sessions over a three-month period in China. Overall, we observed a 1.6% reduction in rebuffering duration and a 6.2% decrease in rebuffering count. At the provincial level11A province in China is similar to a state in the USA., the improvements were even more substantial, with up to a 7.8% reduction in rebuffering duration and a 13.7% decrease in rebuffering count.

With the rapid deployment of LTE/5 G services, mobile subscribers now have access to high-speed services approaching Gbps. However, most mobile data plans have data quota from a few GBs up, beyond which the subscriber will be restricted to much lower bandwidth (e.g., 1 Mbps)-rate-limited service. Rate limiting not only poses a significant challenge to service providers, as it is often mistaken for network problems, triggering false alarms at the providers, but may also cause significant performance anomalies at the application layer and transport layer. This work tackles two central problems in mobile network rate-limiting, namely rate-limiting classification and parameter estimation, through a novel model-based online rate-limiter (MODRL) detector that can detect the presence of rate limiting and estimate its parameters passively from transport layer ACK. Experiments in controlled network testbed and production 4 G/5 G mobile networks show that MODRL can achieve remarkably high and consistent classification accuracy across a wide range of networks. Preliminary results from integrating MODRL into adaptive video streaming and QUIC transport demonstrate that it can effectively eliminate the performance anomalies caused by rate limiting, and open new avenues to further optimize protocol performance over rate-limited mobile networks.

With the rapid advances in high-speed mobile networks such as 5G, Wi-Fi 6, and the upcoming 6G and Wi-Fi 7, streaming live video has become ubiquitous for mobile users. However, live video is susceptible to short-term network condition fluctuations which could lead to video stalls. Our investigations revealed that a substantial portion of such fluctuations were in fact caused by RTT spikes that were not congestion-related. These often confuse the transport protocol into dropping the transmission rate significantly, resulting in video stalls. This motivated us to develop a novel scheme called CWnd-loan to reduce the sender's CWnd-limited idle time during RTT spikes. We applied CWnd-loan to the QUIC protocol with BBR/CUBIC congestion control and strategically deployed it in a tier-1 live video service. The results show that CWnd-loan can effectively reduce sender CWnd-limited idle time by up to 18%, consequently reducing the duration and number of live video stalls by as much as 8.9% and 10.6%. Furthermore, CWnd-loan can also reduce the first-frame time and the playback failure rate by up to 3.2% and 2.7%, respectively. CWnd-loan is designed to complement existing congestion control algorithms and thus could potentially be applied to current as well as future TCP/QUIC designs to tackle RTT spikes commonly found across mobile and wireless networks.

In spite of the advances in mobile networks, most mobile data plans impose a fixed monthly data quota, beyond which the attainable bandwidth is explicitly limited to a much lower data rate. This rate-limited behavior could degrade TCP performances significantly, as confirmed by a major service provider who observed strong correlation between high packet loss rate and mobile rate limiting. The high packet loss translates directly into increased bandwidth cost which is significant in a large-scale service. This work investigates this problem in two steps. First, we establish the link between the high loss rate observed and mobile network rate limiting through experiments in both controlled testbed and production mobile networks. The results revealed that packet loss can and does increase dramatically in rate-limited mobile networks. This affects both TCP Cubic and BBR, the two most widely deployed TCP implementations. BBR, in particular, was impacted far more significantly, resulting in packet loss rates exceeding 40% in some cases. Second, we analyzed BBR's operations under rate limiting to uncover the causes and developed new mechanisms - mBBR, to improve its performance. Experimental results show that mBBR can reduce BBR's packet loss rate by up to 88%, thereby saving substantial bandwidth costs incurred in retransmitting lost packets when the user is under rate limiting.

Learning-based approaches have shown their great potential to adapt themselves to various environments (e.g., PCC and Sprout). Unfortunately, they do not consistently achieve superior QoS across different network conditions and configurations in mobile networks. Furthermore, although they can offer multiple application objectives by adjusting a preference weight vector, it is challenging for users to accurately express an application objective with a weight vector. In this work, we argue that, if configured correctly, the delay-based TCP scheme can outperform learned ones, and allow users to directly specify their objectives. To this end, we propose Post-QoS Analysis (PQSA), a data-driven framework that trains the key QoS-impacting parameters of the scheme to capture the statistical correlations between QoS objectives, network conditions, and configurations, thereby determining the optimal parameter-set that meets the user-defined QoS objective under different network conditions and configurations. To support this, we enhance conventional delay-based TCP design to develop a Generalized TCP-like Rate controller (GR) by exporting three key parameters. Extensive evaluations show that PQSA-optimized GR outperforms existing schemes in different scenarios consistently, and enables service providers to control the QoS flexibly.

Short-video sharing services have seen explosive growth in recent years. Compared to conventional video sharing platforms, these have very different characteristics which are far from well-understood. This work aims at filling the gap by measuring and analyzing detailedapplication-levelperformance data from a top-10 short video service in China. The application-level data offered detailed and rare insights into many performance metrics of the service, which are otherwise inaccessible to external measurements. The service has a scale of over one billion daily views just for the mobile and wireless segments of the service. Our datasets covered over 22 billion video playbacks, over 100 million video files, served by over 5,000 servers to users across 35 provinces and 13 ISPs in China. We analyzed three aspects of the service: (a) video content characteristics; (b) network analytics; and (c) video streaming analytics. Our results revealed significant differences from conventional video-sharing platforms. These findings will have implications for system designs at all levels. The data also enabled us to conduct an indirect network performance measurement of mobile and wireless network services across China,as experiencedby the service. These results offer rare insights into mobile and wireless networks' real-world performance in a large country.

Mobile video streaming is now ubiquitous among mobile users. This work investigates a less studied and yet significant problem in mobile video streaming – data wastage, i.e., some downloaded video data may not be played back but discarded by video players due to early departure or video skip, thus the bandwidth consumed in transferring them is wasted. Our measurements show that data wastage is significant in practice, e.g., 25.2 percent∼51.7 percent of video data downloaded are in fact wasted. Moreover, substantial data wastage exists not only in current commercial streaming platforms, but also in state-of-the-art adaptive streaming systems proposed in the literature. This work develops a new post-streaming wastage analysis (PSWA) framework to tackle this problem by converting existing adaptive streaming algorithms into data wastage aware versions. PSWA enables the streaming vendors to explicitly control the tradeoff between data wastage and quality-of-experience (QoE). Extensive evaluations show that PSWA can reduce data wastage significantly (e.g., 80 percent) without any adverse impact on QoE. Moreover, it has strong robustness to perform consistently across a wide range of networks. PSWA can be readily implemented into current streaming platforms, and thus offers a practical solution to data wastage for mobile streaming services.

Video streaming has grown tremendously in recent years and it is now one of the main applications on the Internet. Due to the networks' inherent bandwidth fluctuations, various rate-adaptive streaming algorithms have been developed to compensate for such fluctuations to improve Quality-of-Experience (QoE). However, in practice, the preference for QoE typically differs significantly across different viewers and there is no systematic way so far to comprehensively incorporate different sets of conflicting QoE objectives into the algorithm design. Thus, it is not surprising that the QoE performance achieved by the existing algorithms is in fact far from optimal. This work aims at attacking the heart of the problem by developing a novel framework called Post Streaming Quality Analysis (PSQA) that can maximize the QoE under any preference through automatically tuning the adaptation logic of the streaming algorithms. Evaluation results show that the QoE achieved by PSQA is substantially better than the existing approaches and in some scenarios even close to optimal. Moreover, PSQA can be readily implemented into real streaming platforms, offering a practical and reliable solution for high-performance streaming services.

Fueled by the emerging short video applications (e.g., TikTok), streaming short-form videos nowadays is ubiquitous among mobile users. During the viewing, one common action is to scroll the screen to switch videos, which is a handy operation for the viewers to quickly search for content of interest. However, our empirical measurements reveal that frequent video switching can result in nearly half of the mobile data quota being used for transferring the video data that is never watched. This problem is called data loss in this work. Given the immense cost of the network infrastructure, such a high proportion of data loss is financially tremendous to both mobile users and streaming vendors. To tackle the problem, this study proposes a novel system called Data Usage Aware Short Video Streaming (DUASVS), where a new Integrated Learning is used to capture the characters of past network conditions and then trains intelligent adaptation models to reduce data loss and save data usage. Extensive evaluations show that DUASVS is able to save 70.7%∼83.2% of mobile data usage without incurring any QoE degradation. Moreover, the system exhibits strong robustness, performing consistently over a wide range of network environments as well as video streaming sessions.

With the progressive deployment of 5G networks around the world, mobile networks are entering a new era where bandwidth will be breaking through the Gbps barrier. In this work, we investigate the performance of current TCP designs in such high-bandwidth networks, demonstrating the potential bottleneck due to TCP’s Slow-Start mechanism which is an integral component in most TCP designs. For example, transferring a file of 1 MB size in a first-generation 5G network using Linux’s default TCP-Cubic and Google’s TCP-BBR resulted in average throughputs of 18.2 Mbps and 32.8 Mbps, respectively. Compared to the mean available bandwidth of 180 Mbps, the gap is significant. To tackle this problem, we developed an enhanced Stateful-TCP technique to transform BBR into a new S-BBR to accelerate its startup performance to narrow the gap. Results from trace-driven emulated 5G network experiments show that S-BBR could improve BBR’s throughput performance by 50% to 100% while maintaining similar delay performance. This is further validated by an independent competitive benchmark using over 500 clients where S-BBR raised BBR’s throughput by 69%. S-BBR is sender-based and thus can be readily deployed in Internet servers without any requirements from the client side, it retains BBR’s desirable features and so offers a promising solution to enhance mobile applications’ performance in the emerging high-bandwidth mobile and wireless networks.

Live video streaming has seen tremendous growth in the past decade. An important fact in live streaming is that the demand for low playback-latency inherently conflicts with the desire for high QoE. This requires different types of live services to seek different latency-QoE tradeoffs according to their service-requirements. However, our investigations revealed that it is fundamentally difficult for existing streaming algorithms to keep consistent latency in changing network conditions, let alone achieve the service-desired latency-QoE tradeoff. To tackle the challenge, this article develops a novel framework called Flexible Latency Aware Streaming (FLAS) that not only can achieve consistent low latency, but also control the latency-QoE tradeoff flexibly. Specifically, FLAS generates a set of adaptation logics offline, each optimized for a candidate tradeoff point, then selects the most appropriate one to run online. We first show how FLAS can be applied to optimizing the existing algorithms, then developed a novel Genetic Programming approach to fully exploit FLAS's potential. Extensive evaluations show that FLAS can precisely control latency all the way down to 1s and achieve substantially higher QoE than state-of-the-arts. FLAS can be readily implemented into real streaming platforms, offering a practical and reliable solution for live-streaming services.

Video streaming is now ubiquitous in the mobile Internet. This motivated intense research in adaptive streaming algorithms to tackle mobile networks' fluctuating conditions. Our investigations revealed that while existing algorithms can perform well in their intended operating environments, their performance can degrade substantially in other environments. This work tackles this challenge by developing a novel Ensemble Adaptive Streaming (EAS) paradigm to mobile video streaming. As opposed to designing a single streaming algorithm for all network conditions, we argue that different network conditions require different algorithms. We introduce the notion of network differentiators to segregate network conditions into different classes where each class has its own adaptation algorithm designed and optimized specifically for it. An EAS mobile streaming client then selects at runtime the matching adaptation algorithm using the same network differentiators on a per session basis for streaming. We show how EAS can be applied to existing machine-learning approaches to improve their performances. Moreover, to fully exploit EAS's potential we developed the first Genetic Programming approach to evolve adaptive streaming algorithms. The resultant EAS-GP algorithms not only outperformed state-of-the-art algorithms substantially, but also exhibited remarkable robustness over time, location, mobile operators, as well as quality-of-experience metrics.

Live video streaming is one of the fastest growing applications in the mobile Internet. In addition to video quality and streaming performance, live video typically requires shorter user viewing latency or else the service quality will degrade significantly (e.g., live sports). This poses a significant challenge as latency and quality-of-experience (QoE) are inherently conflicting objectives. This work tackles this challenge by developing a novel LAPAS algorithm that employs playback rate adaptation in addition to bitrate adaptation to: (a) enable the service provider to explicitly control the tradeoff between latency and QoE; (b) reduce latency while maintaining reasonably high QoE. Extensive trace-driven simulation results showed that LAPAS can reduce latency by 21.8% to 75.1% at the same or higher QoE compared to leading algorithms. Moreover, it is the only algorithm that can achieve viewing latency shorter than 4 seconds and can work well even for latency as short as 2 seconds. Last but not least, LAPAS's performance is robust with respect to network locations, mobile operators, and time. It can be implemented entirely at the client side and is thus readily deployable in today's mobile Internet.

Small-cell architecture is widely adopted by cellular network operators to increase spectral spatial efficiency. However, this approach suffers from low spectrum temporal efficiency. When a cell becomes smaller and covers fewer users, its total traffic fluctuates significantly due to insufficient traffic aggregation and exhibits a large “peak-to-mean” ratio. As operators customarily provision spectrum for peak traffic, large traffic temporal fluctuation inevitably leads to low spectrum temporal efficiency. To address this issue, in this paper, we advocate device-to-device (D2D) load-balancing as a useful mechanism. The idea is to shift traffic from a congested cell to its adjacent under-utilized cells by leveraging inter-cell D2D communication, so that the traffic can be served without using extra spectrum, effectively improving the spectrum temporal efficiency. We provide theoretical modeling and analysis to characterize the benefit of D2D load balancing, in terms of total spectrum requirements and the corresponding cost, in terms of incurred D2D traffic overhead. We carry out empirical evaluations based on real-world 4G data traces and show that D2D load balancing can reduce the spectrum requirement by 25% as compared to the standard scenario without D2D load balancing, at the expense of negligible 0.7% D2D traffic overhead.

Mobile video streaming is now ubiquitous among mobile users. This work investigated an often-neglected problem - data wastage where downloaded video data were not played back due to user early departure. Empirical measurements showed that data wastage is significant, e.g., around 20% of data downloaded were in fact wasted. Moreover, substantial data wastage exists not only in current commercial streaming platforms, but also in advanced adaptive streaming algorithms proposed in the literature. This work developed a new Post-Streaming Wastage Analysis (PSWA) framework to tackle the problem by converting existing adaptive streaming algorithms into wastage-aware. PSWA enables the service provider to control the tradeoff between data wastage and streaming quality-of-experience (QoE). Most remarkably, PSWA can achieve significant data wastage reduction (e.g., over 70%) even without negatively impacting QoE. PSWA can be applied to existing or future adaptive streaming algorithms and thus offers a practical solution to data wastage in current and future streaming services.

Fueled by the growth of 3G/4G mobile networks, mobile video streaming has become one of the main applications in the mobile Internet. Due to mobile networks' inherent bandwidth fluctuations, the industry as well as researchers have developed many adaptive streaming algorithms to compensate for such fluctuations to improve streaming performance. Given the wide range of network settings, it is not surprising that existing algorithms can and do perform differently across different network and system conditions. This work breaks away from the conventional one-size-fits-all approach to designing adaptive streaming systems by developing a new framework called PSRA where past throughput trace data - captured as a by-product of streaming, are analyzed to construct a statistical model to automatically tune the adaptation algorithm for future streaming sessions according to the underlying network and system configurations. Compared to existing approaches, the PSRA-optimized streaming algorithm can achieve predictable, consistent, and controllable streaming performance across a wide-range of network and system configurations. Moreover, PSRA offers to service provider a new tool to precisely control the tradeoff between video quality and streaming performance. Results from extensive trace-driven simulations as well as experiments verified PSRA's performance under real-world mobile network and system configurations.

Today's datacenter is shared among various applications with different QoS requirements, which poses a great challenge to deliver low delay transport with high throughput. Most of works address this challenge by reducing the in-network delay, but assumes a negligible local delay. However, we show that this assumption does not hold for a multi-tenant datacenter that a physical machine is shared by multiple tenants with virtual machines running different applications. As measured, we found that VMs in a PM competing for bandwidth resources introduce delays as high as 13 ms, resulted from the packet queueing at QDisc layer of that PM, because current VMs' rate control still operates in a distributed manner without exploiting knowledge of the QoS requirements of applications running in VMs. This work addresses this problem by proposing a centralized rate adaptation (CERA) that operates in the host PM, dynamically schedules the flows from all VMs in a centralized manner. We implemented a CERA prototype and evaluated CERA through testbed experiments. Our results show that CERA reduces the local delay significantly thus reduces the average request latency of delay sensitive applications, e.g., memcached, by a factor of 6.3, without sacrificing the throughput performance of throughput intensive applications, e.g., iperf.

Mobile video streaming is one of the fastest growing applications in the mobile Internet. Nevertheless, delivering high-quality streaming video over mobile networks remains a challenge. Researchers have since developed various novel streaming algorithms such as rate-adaptive streaming to improve the performance of mobile streaming services. However, selection or optimization of streaming algorithms is far from trivial and there is no systematic way to incorporate the tradeoffs between various performance metrics. This work aims at attacking the heart of the problem by developing a novel framework called Post Streaming Quality Analysis (PSQA) to automatically tune any streaming algorithms to maximize a given quality-of-experience (QoE) objective. We show that PSQA not only can be applied to optimize the performance of existing streaming algorithms, but also opens a new way for the exploration of new adaptive video streaming protocols and QoE metrics. Simulation results based on real network throughput traces show that PSQA can optimize existing and new streaming algorithms to achieve QoE that is remarkably close to the optimal achieved using brute-force method ex post facto.

Mobile data networks exhibit highly variable data rates and stochastic non-congestion-related packet loss. These challenges result in key performance bottlenecks in current Transmission Control Protocol (TCP) implementations: bandwidth inefficiency and large end-to-end delay. This work addresses these challenges by first developing a Sliding Interval based Rate Adaptation (SIRA) that tracks bandwidths with a fixed time interval and applies them to its transmission rate periodically. Extensive experiments confirmed that SIRA achieves 96.3% bandwidth utilization and reduces the average queueing delay by a factor of 1.37, compared to TCP CUBIC, the preferred variant for Internet servers. However, the resultant end-to-end delay is still much larger for interactive applications, thus we complement SIRA with two heuristic rate compensation algorithms (SIRA-H) given that the bandwidth does not vary significantly in long time scales. Specifically, SIRA-H first reduces the transmission rate of SIRA if the estimated RTT is above a prefigured threshold. Meanwhile, it computes the amount of unsent data that would be transmitted if SIRA were used, and compensates the rate reduction with those unsent data as if their ACKs were received, when the queue is detected to be empty. We evaluated SIRA-H through a combination of trace-driven emulations and real-world experiments, and showed that it reduces the 95thpercentile queueing delay by a factor of over 3.9, while maintains a similar throughput compared to the original SIRA. In comparison to state of the art protocols such as Sprout and Verus, SIRA-H also reduces the 95thpercentile queueing delay by a factor of over 0.8.

Mobile video streaming services are ubiquitous today and yet their real-world performance is still largely inconsistent, unpredictable, and uncontrollable, primarily due to mobile networks' inherent bandwidth fluctuations at both short and long timescales. With the rapid emergence of paid streaming services, picture and streaming qualities can no longer be an afterthought. This work develops a novel framework called Variable-Bitrate Post-Streaming Rate Analysis (VBR-PSRA) that, for the first time, enables service providers and mobile operators to provision higher-quality VBR-encoded video streaming services with consistent, predictable, and controllable performance. The VBR-PSRA framework exploits past throughput trace data collected during actual streaming sessions to construct a statistical model that captures and quantifies the relation between recent throughput data, video quality choice, video bitrate variations, and streaming performance. It allows the service provider to set a target streaming performance in terms of playback rebuffering probability where it will then automatically select the best video quality that can be streamed. Extensive simulations using trace data obtained from production 3G/HSPA networks in three different locations showed that the proposed VBR-PSRA framework can achieve actual streaming performances which are remarkably close to the target.

Mobile data networks such as 3G and LTE exhibit properties that are fundamentally different from those of fixed networks. These differences result in severe performance bottlenecks in current Transmission Control Protocol (TCP) implementations, which are the foundation for most of today's mobile applications. This work addresses this challenge by developing a transparent protocol optimization device to perform on-the-fly protocol optimization to improve TCP's throughput performance while maintaining full compatibility with current end-host TCP implementations. The proposed protocol optimizations can achieve near optimal bandwidth utilization. This was verified and confirmed in production 3G and LTE networks using a prototype implementation. Compared to current TCP implementations, the proposed protocol optimization device can raise throughput by 48 to 163 percent. In contrast to inventing a new transport protocol or modifying an existing TCP implementation, the proposed approach does not require any modification to the existing TCP implementation at the client/server hosts, does not require any reconfiguration of the server or client, and hence can be readily deployed in today's 3G and 4G mobile networks, raising the throughput performance of all existing network applications running atop TCP.

Bandwidth-sensitive applications such as adaptive video streaming rely on accurate prediction of future network throughput to enable them to react to and compensate for the rapidly fluctuating bandwidth often found in mobile networks. Researchers have developed various prediction algorithms in the literature of which many have been employed in real-world applications. However, there is a lack of systematic study on the comparative performance of the existing prediction algorithms in the context of mobile networks. This work addresses this void by conducting a systematic performance comparison of 7 prediction algorithms, and analyzes their characteristics when applied to the prediction of TCP throughput in mobile networks. The performance results are obtained from extensive trace-driven simulations where the throughput trace data were captured in production 3G/HSPA mobile networks in 3 locations over a period of 9 months and hence offer a good representation of the prediction algorithms' real- world performance. Furthermore, we applied the theory of differential entropy in information theory to obtain an estimated lower bound on throughput prediction errors which, for the first time, enables one to evaluate the absolute performance of these prediction algorithms. The results revealed that more complex algorithms are not necessarily better, and there exists a specific range of operating parameters where predictions are generally more accurate.

Small-cell architecture is widely adopted by cellular network operators to increase network capacity. By reducing the size of cells, operators can pack more (low-power) base stations in an area to better serve the growing demands, without causing extra interference. However, this approach suffers from low spectrum temporal efficiency. When a cell becomes smaller and covers fewer users, its total traffic fluctuates significantly due to insufficient traffic aggregation and exhibiting a large "peak to-mean" ratio. As operators customarily provision spectrum for peak traffic, large traffic temporal fluctuation inevitably leads to low spectrum temporal efficiency. In this work, we first carryout a case-study based on real-world 3G data traffic traces and confirm that 90% of the cells in a metropolitan district are less than 40% utilized. Our study also reveals that peak traffic of adjacent cells are highly asynchronous. Motivated by these observations, we advocate device-to-device (D2D) load-balancing as a useful mechanism to address the fundamental drawback of small-cell architecture. The idea is to shift traffic from a congested cell to its adjacent under-utilized cells by leveraging inter-cell D2D communication, so that the traffic can be served without using extra spectrum, effectively improving the spectrum temporal efficiency. We provide theoretical modeling and analysis to characterize the benefit of D2D load balancing, in terms of sum peak traffic reduction of individual cells. We also derive the corresponding cost, in terms of incurred D2D traffic overhead. We carry out empirical evaluations based on real-world 3G data traces to gauge the benefit and cost of D2D load balancing under practical settings. The results show that D2D load balancing can reduce the sum peak traffic of individual cells by 35% as compared to the standard scenario without D2D load balancing, at the expense of 45% D2D traffic overhead.

Mobile video streaming has become a mainstream application due to vastly improved smartphone hardware and mobile network capacity in recent years. Nevertheless, mobile video streaming remains challenging in practice due to mobile network's inherent bandwidth fluctuations. This work tackles two long-standing challenges in mobile video streaming, namely to provision streaming services with predictable streaming performance and guaranteed video quality. In contrast to existing approaches based on adaptive video streaming, we show that today's mobile networks, despite the seemingly random bandwidth fluctuations, do exhibit statistically significant correlations over short and long time-scales. By developing a new framework to correlate the statistical correlations between video bitrate, streaming performance, and startup delay, we show that it is both possible and practical to achieve the above two goals by adaptively configuring the startup delay. Trace-driven simulations based on bandwidth traces captured from production 3G networks show that the proposed framework can readily achieve both streaming performance and video quality guarantees in today's mobile networks.

Adaptive video streaming is an essential tool for improving the performance of video delivery over mobile networks. By dynamically switching between different bit-rate versions of the same video, adaptive video streaming can compensate for and adapt to the ever-changing network conditions inherent in today's 3G/4G networks. However, existing adaptive streaming algorithms, both academic and commercial ones, do not offer any prediction on the streaming performance of future streaming sessions, nor allow the content providers or users to explicitly control the tradeoff between streaming performance and video quality. This study tackles this fundamental challenge by developing a novel framework called Throughput-Differentiated-Rate-Adaptive Post-Streaming-Rate-Analysis (TDRA-PSRA) based on a new statistical model that directly relates past bandwidth statistics to future streaming performance. Extensive simulation results obtained from real-world mobile network trace data revealed three remarkable properties of TDRA-PSRA: (a) the actual average streaming performance is very close to the target set forth by the ICPs/users; (b) it enables the ICPs/users to control the tradeoff between streaming performance and video quality; (c) it offers a mean to directly control the frequency of bit-rate switches — a key factor to subjective video quality.

Much work has been done to improve the performance of video streaming over mobile data networks. The widespread adoption of HTTP/TCP for video streaming further complicates the problem as TCP's own dynamics add even more fluctuations to the already unpredictable network bandwidth. Not surprisingly, none of the existing video bit-rate selection algorithms can achieve consistent or predictable streaming performance in the presence of wide network bandwidth fluctuations. This work tackles this problem by developing a new model to capture the statistical correlations between streaming performance, video bit-rate selection, and past TCP throughput data. Based on a novel post-streaming rate analysis technique, the proposed model can be used to predict the streaming performance of future video sessions at any given video bit-rate choices. This not only enables content providers to control the quality-of-service for streaming users, but also to ensure consistent streaming performance for users with widely different access bandwidths.

Knowledge of the queue length and link buffer size of the bottleneck link has many potential applications such as congestion control, traffic engineering, traffic policing, content adaptation, QoS monitoring and provisioning, etc. A recent work proposed a new Sum-of-Delay (SoD) algorithm which can accurately estimate such link properties in mobile networks. This work presents a delay variation compensation algorithm - Sum-of-Delay with Timestamp (SoD-TS), to further improve SoD's estimation accuracy in networks with large uplink delay variations. The proposed algorithm exploits the existing TCP Timestamp option which is widely deployed and thus can be implemented without modification to the receiver's TCP implementation. By making use of SoD-TS we developed a novel transport layer application - queue-length-based congestion control algorithm (QCC), on top of TCP that tackles the problem of bufferbloat to provide better QoS for applications requiring low end-to-end delay and/or high bandwidth utilization. Trace-driven simulation shows that SoD-TS can effectively eliminate estimation errors caused by uplink delay variations. Moreover, by applying QCC to the widely deployed TCP CUBIC it can reduce the RTT by a factor of over 3.6 while still achieving over 87% bandwidth utilization.

Knowledge of the queue length of the bottleneck link has many potential applications such as congestion control, traffic engineering, traffic policing, content adaptation, QoS monitoring and provisioning, etc. A recent work proposed a new Sum-of-Delay with Timestamp (SoD-TS) algorithm which can accurately estimate queue length in mobile networks with both bandwidth variations and uplink delay variations by exploiting the existing TCP Timestamp option. By making use of SoD-TS we developed a novel transport protocol - queue-length-aware TCP (TCP-QLA), that tackles the problem of bufferbloat to provide better QoS for applications requiring low end-to-end delay and/or high bandwidth utilization. Trace-driven simulation shows that compared to TCP CUBIC TCP QLA can reduce the RTT by a factor of 2.7 while still achieving over 97% bandwidth utilization. Moreover, TCP-QLA further reduces RTT by 50% compared to delay-based TCP such as FAST TCP and TCP Vegas.

The emerging mobile data networks fueled by the world-wide deployment of 3G, HSPA, and LTE networks created new challenges for the development of Internet applications. Unlike their wired counterpart, mobile data networks are known to exhibit highly variable bandwidth. Moreover, base stations are often equipped with large buffers to absorb bandwidth fluctuations to prevent unnecessary packet losses. Consequently to optimize protocol performance in mobile data networks it is essential to be able to accurately characterize two key network properties: queue length and buffer size of the bottleneck link. This work tackles the challenge in estimating these two network properties in modern mobile data networks. Using extensive trace-driven simulations based on actual bandwidth trace data measured from production mobile data networks, we show that existing queue-length and link buffer size estimation algorithms no longer work well in bandwidth-varying networks. We develop a novel sum-of-delays algorithm which incorporates the effect of bandwidth variations into its estimation. Extensive trace-driven simulation results show that it can accurately estimate the queue length and link buffer size under both fixed and varying bandwidth conditions, outperforming existing algorithms by up to two orders of magnitude.

Recent advances in high-speed mobile networks have revealed new bottlenecks in the ubiquitous TCP protocol deployed in the Internet. In addition to differentiating random loss from congestion loss, our experiments revealed that TCP's flow control mechanism can become a significant bottleneck during TCP's loss recovery phase, resulting in bandwidth efficiency as low as 0.05. To tackle this problem we develop a novel opportunistic retransmission algorithm to enable the TCP sender to transmit new packets even before the loss recovery phase is completed, resulting in significant improvement in TCP's bandwidth efficiency. We applied the proposed mechanism to three TCP variants and developed system models to analyze their performance. Experimental results showed that TCP's throughput can be improved by up to 56% in real-world settings.

This paper investigates the performance of TCP in mobile data networks and proposes a novel approach to address the problem of optimizing transport protocol for these networks - a network-centric approach. We propose to realize protocol optimizations within the network by means of deploying a network-layer device - called a mobile accelerator, which performs protocol optimizations on-the-fly to improve transport protocol performance, without requiring any modification to the server or the client protocol implementations in the OS. Our extensive experiments conducted in production 3G/HSPA and LTE networks show that the proposed mobile accelerator can increase the throughput performance of TCP by over 200%. This mobile accelerator can be readily deployed in existing mobile data networks and can improve the network performance of all existing network applications running atop TCP.

The rapidly emerging mobile data networks fueled by the world-wide deployment of 3G, HSPA, and LTE networks created new challenges for the development of Internet applications. Unlike their wired counterpart, mobile data networks are known to exhibit highly variable bandwidth. Moreover, base stations are often equipped with large buffers to absorb bandwidth fluctuations to prevent unnecessary packet losses. However these two factors together invalidate the assumptions used in conventional protocol designs, which routers are assumed to have fixed link capacity and a small buffer. Consequently to optimize protocol performance in mobile data networks it is essential to be able to accurately characterize the network properties such as bottleneck link buffer size. This work tackles the challenge in estimating link buffer size in modern mobile data networks. Using extensive trace-driven simulations based on actual bandwidth trace data measured from production mobile data networks, we show that existing link buffer size estimation algorithms such as max-min and loss-pair no longer work well in bandwidth-varying networks. We develop a novel sum-of-delays algorithm to tackle this challenge. Simulation results show that it can accurately estimate the link buffer size under both fixed and varying bandwidth conditions, outperforming the existing algorithms by multiple orders of magnitude.

The success of peer-to-peer (P2P) applications hinges on users' willingness to contribute their network bandwidth to serve other peers. However if the upload data rate to other peers is too high it could severely degrade the download data rate, even if the downlink has abundant bandwidth available, especially in an asymmetric network such as ADSL. Experiments revealed that the download throughput degradation is in fact not caused by congestion in the uplink, but caused by increased queuing delay in the uplink path during high upload data rates. This paper tackles this problem by developing an adaptive algorithm to monitor the uplink queuing delay and adjust the upload data rate limit dynamically so that the download throughput will not be adversely affected. Experiments conducted using an open-source P2P software showed that the proposed algorithms can increase the downlink utilization over a wide range of network configurations (and over 200% increase in some cases) by automatically adjusting the upload data rate limit. The algorithms do not require any user intervention and can be readily incorporated into existing P2P systems.

The efficient scheduling of streaming data delivery in a peer-to-peer (P2P) network is a hard problem due to the Internet's lack of support for resource allocation and performance guarantees. In particular, the bandwidth resources available to a peer is constantly in flux and the future bandwidth availability is very difficult, if not impossible, to predict accurately. This work proposes to tackle this problem from a different angle. We investigate the use of erasure codes to encode the media data and then schedule multiple peers to stream the encoded data simultaneously to a receiver. By exploiting the order-invariant property of erasure codes this approach enables the sending peers to fully utilize their available bandwidth resources and yet does not need to estimate or predict their bandwidth availability. Moreover, we develop distributed scheduling algorithms to juxtapose the data transmissions from multiple peers so that the coding and storage complexities can be kept at practical level in scaling up the system. This paper describes the motivation, architecture, and design of the proposed coding/scheduling algorithms; develops a performance model to characterize the algorithms' performance bounds; and evaluates them through simulation as well as experiments.

Application-layer multicast (ALM) has grown tremendously in recent years, making the distribution of bulk data such as streaming video economically feasible for small companies and even individuals. The efficiency of an ALM network depends on its data distribution overlay, which is constructed based on metrics such as round-trip time (RTT) measurement between peers. However Internet measurement experiments revealed that RTT is far from an accurate estimator of bandwidth availability and as such, may lead to sub-optimal performance in the constructed ALM overlays. This paper tackles this problem by developing a new in-band bandwidth probing tool which can estimate the amount of achievable bandwidth available in the target network path so that excess data traffic can be diverted from the congested path without causing new congestion in the target path. Moreover, the probing tool does not incur any bandwidth overhead as it piggybacks on the existing data flow. Simulation results show that multioverlay ALM networks constructed based on achievable bandwidth consistently out-performs RTT-based and residual bandwidth-based approaches in terms of data delivery ratio and video playback continuity. Moreover, the proposed bandwidth probing tool can be implemented entirely within the application and thus can be readily incorporated into existing ALM protocols.

Recent advances in multicast video streaming algorithms have opened up new ways to provision video-on-demand services to potentially millions of users. However, the spectacular efficiency of multicast streaming algorithms can only be realized by restricting or even prohibiting interactive playback control. Experiments reveal that the performance of current state-of-the-art multicast streaming algorithms will degrade significantly even at very low levels of interactivity (e.g., one control per five users). This study tackles this challenge by investigating the fundamental limitations of multicast streaming algorithms in supporting interactive playback control and presents a general solution-static full stream scheduling (SFSS)-which can be applied to many of the existing multicast streaming algorithms to substantially improve their performance when interactive playback control is to be supported. Moreover, to solve the problem of optimizing the algorithm for the often unknown client access patterns (e.g., arrival rates and interactivity rates), we present a novel just-in-time simulation (JTS) scheme to dynamically and automatically tune operating parameters of the SFSS algorithm while the system is online. This JTS scheme not only eliminates the need for a priori knowledge of the often unknown system parameters, but also can adapt to changes in the client access pattern over time. Extensive simulation results show that the proposed adaptive algorithm can reduce the admission and interactive control latencies by as much as 90%

The current best-effort Internet does not guarantee the bandwidth availability between a receiver and a sender, and so renders any quality-of-service control difficult, if not impossible. This paper proposes a novel predictive buffering algorithm for streaming video not from one, but from multiple senders to a receiver over the best-effort Internet. In particular, the predictive buffering algorithm estimates the mean and variance of the aggregate throughput of multiple senders, and then use these estimated parameters to predict the future bandwidth availability. By appealing to the central limit theorem, the future bandwidth availability will tend to be normally distributed, irrespective of the distribution of the measurement bandwidth availability. This insight enables the buffering algorithm to predict, at runtime, the buffering time required to ensure playback continuity. Extensive trace-driven simulations show that this predictive buffering algorithm can achieve buffer delays that are remarkably close to the optimal buffer time.

In recent years, a number of sophisticated architectures have been proposed to provide video-on-demand (VoD) service using multicast transmissions. Compared to their unicast counterparts, these multicast VoD systems are highly scalable and can potentially serve millions of concurrent users. Nevertheless, these systems are designed for streaming constant-bit rate (CBR) encoded videos and thus cannot benefit from the improved visual quality obtainable from variable-bit rate (VBR) encoding techniques. To tackle this challenge, this paper presents a turbo-slice-and-patch (TSP) algorithm to support VBR video streaming in a multicast VoD system. Results obtained from trace-driven simulation of 300 VBR videos show that serving VBR videos with the TSP algorithm increases the average latency by only 9% compared to the CBR case with the same average video bit rate. Moreover, in 165 out of the 300 video titles, the TSP algorithm actually outperforms the CBR equivalent by shortening the latency by 0.04%-99%. Given that we can achieve similar visual quality by encoding VBR video at half the average rate of CBR video, this TSP algorithm can potentially serve VBR videos with more consistent visual quality and with less resource compare to CBR-based video streaming systems.

Recently, a server-less architecture has been proposed for building video streaming systems which does not need any costly dedicated video servers and yet is highly scalable and reliable. However, due to the potentially large. number of user hosts streaming video data to a receiver for playback, the aggregate network traffic can become very bursty, leading to significant packet loss at the access routers. This study tackles this problem by investigating a novel network-aware transmission scheduling algorithm called gradient-descent scheduler (GDS) to reduce the traffic burstiness. Simulation results demonstrate that GDS can reduce the congestion-induced packet loss from over 95% to 0.07% in a 500-host system. Moreover, GDS can automatically adapt to the underlying network and does not require hosts in the system to be synchronized. These are essential for practical design of server-less architectures and peer-to-peer systems.

Periodic broadcasting (PB) schemes are the most promising solution for building large-scale video streaming services. Existing PB schemes are all built around the traditional client-server model. This paper proposes a radically different architecture where only end-user hosts are used to build the PB system, thereby eliminating the need for a central server or even a service provider. Two unique problems pertaining to this decentralized, peer-to-peer architecture is addressed: transmission scheduling and peer synchronization. A new decentralized synchronization algorithm is proposed and results obtained from simulations confirm the architecture's feasibility and performance.

The concept of care-of-address (CoA) is a major cause of excessive handoff delay in Mobile IPv6 for real time multimedia traffic. Many schemes eliminate the use of CoA at the micro-mobility scale, but leave the macro-mobility unsolved. This paper proposes a novel alternative IPv6 mobility scheme based on universal addressing - Cellular Universal IP (CUIP) - for real-time traffic in wireless access networks. In CUIP, a mobile node is addressed with a universal IP address regardless of its location, making CoA and tunneling unnecessary in micromobility and even macromobility handoffs. CUIP manages roaming and handoff differently - whereas explicit signaling is used for roaming, a handoff-on-the-fly route-update scheme is used during handoff to embed signaling information into the outgoing data packets to minimize handoff delay. We prove analytically that, on average, fewer than three routers need to be updated per handoff. As a result, CUIP incurs an expected network layer handoff delay on the order of milliseconds only. In addition, the support of QoS is possible. A simple security scheme is also proposed to enable mutual authentication at the network layer.

Variable-bit-rate (VBR) encoded videos can provide a more consistent visual quality than constant-bit-rate (CBR) encoded videos. However, the long-range bit-rate variations in VBR videos make it difficult to efficiently provide quality-of-service control in a video-on-demand system. Existing scheduling algorithms such as Optimal Smoothing, which requires both downward and upward bandwidth reallocations to adapt to the video bit-rate variations, simply cannot guarantee video delivery in networks with mixed video and data traffic. This study tackles this limitation by investigating a new scheduling algorithm with monotonic-decreasing rate allocations for scheduling video data transmissions. By eliminating upward bandwidth reallocations, the proposed scheduler can guarantee video delivery even in the presence of other data traffic in the same network. Moreover, results show that the proposed scheduler can achieve such performance guarantee without tradeoff in performance or resource requirements. This work presents this new monotonic-decreasing rate scheduler, analyzes its fundamental properties, and evaluates its performance using a large number of real-world VBR video traces (274 DVD movies) in extensive trace-driven simulations.

Recently a number of researchers have proposed and investigated new video-on-demand architectures that make use of network multicast to achieve vastly improved efficiency. Techniques such as batching, patching, periodic broadcasting, chaining, and piggybacking have been explored both in isolation and in combinations. This study investigates a new tool in the arsenal-channel folding, where aggressive client-side caching is used to merge clients from one multicast channel into the other. In particular, this channel folding algorithm is applied to a previously proposed unified video-on-demand (UVoD)) architecture to demonstrate and to quantify its impact on the performance and the tradeoff in a multicast video distribution architecture. This paper presents this channel folding algorithm in the context of UVoD) and derives a performance model to obtain the system latency, the near-optimal channel partition policy, and the client buffer requirement. Numerical results show that channel folding can double the capacity of UVoD) with a remarkably small overhead in the client buffer requirement.

The lack of end-to-end quality of service support in the current Internet has caused significant difficulties to ensuring playback continuity in video streaming applications. This study addresses this challenge by investigating a new adaptation algorithm to adjust the bit-rate of video data in response to the network bandwidth available to improve playback continuity. Unlike previous works, the proposed algorithm is transparent to the video client, requires no parameter tuning, and yet can outperform existing algorithms. This paper presents this algorithm, evaluates and compares its performance with the best algorithm currently available using extensive trace-driven simulations.

Recently, a server-less video-on-demand architecture has been proposed to eliminate costly dedicated video servers and yet it is highly scalable and reliable. However, because of the potentially large number of user hosts streaming video data to a receiver for playback, the aggregate network traffic can become very bursty, leading to significant packet loss at the access routers. To tackle this problem, we propose a novel network-aware transmission scheduler called least schedulable first (LSF) to reduce the traffic burstiness. Simulation results show that LSF can reduce the congestion-induced packet loss from over 95% to 0.05% in a 500-host system at 0.95 system utilization. Moreover, LSF can adapt to variations in the underlying network, is inherently distributed, and does not require synchronization among hosts in the system.

The Internet nowadays consists of multicast-capable domains or "islands" interconnected by multicast-incapable routers. In order to achieve efficient global multicast, we propose and study island multicast (IM) where overlay connections are used between islands while IP multicast is used within an island. IM may use any existing application-level multicast protocol to build island overlay. We describe how to elect a representative (or leader) in each island for such a purpose. We also present the mechanisms for electing the bridging nodes for overlay connections. Using Internet-like topologies, we show that IM achieves much higher bandwidth efficiency as compared to using application-level multicast alone, at the cost of a small increase in end-to-end delay.

This paper proposes an integrated routing and grooming algorithm for IP over WDM networks. Assuming a peer model in GMPLS-Based optical networks, we take into account the combined topology and resource usage information on both IP and WDM layers. Based on a clustering technique called blocking island paradigm, we propose an enhanced blocking island graph (BIG) network model with blocking island hierarchy (BIH) to abstract network resources. The main idea of the algorithm is to keep the integrity and load balance of related blocking islands. We also combine a cost function in the routing algorithm to groom traffic flows into active lightpaths. The complexity of the algorithm is analyzed to show its efficiency. In the simulation, we compare the algorithm with three other integrated routing algorithms in terms of blocking probability. The three algorithms are: the integrated min-hop (IMH) routing algorithm, the maximum open capacity (MOCA) routing algorithm and the IP-WDM grooming (IWG) algorithm. Simulation results show our algorithm has the best performance.

This paper proposes and investigates a simple self-adaptive contention window adjustment algorithm for 802.11 WLAN. We present simulation and analytical results showing that the new algorithm outperforms the standard 802.11 window-adjustment algorithm. Compared with the standard and previously proposed enhancement algorithms, a salient feature of our algorithm is that it performs well both when the number of active stations is large and small that is, in both heavy and light contention cases. Furthermore, the adaptive window adjustment algorithm is simpler than previously proposed enhancement schemes in that no live measurement of the WLAN traffic activity is needed.

The rapid growth of wireless LAN (WLAN) deployments will bring about many novel mobile applications. Among them will be real-time multimedia streaming applications running on UDP, which may interfere with current data applications running on TCP. This paper is a first attempt to investigate how to ensure the performance of these two groups of applications when they co-exist over a WLAN. Toward this end, we have designed and implemented a UDP rate adaptation scheme called adaptive-buffer rate control (ABRC) for multimedia streaming over WLAN. ABRC has two distinguishing features compared with other schemes: 1) it can achieve arbitrary bandwidth allocations between UDP and TCP in the WLAN, as opposed Io previously proposed "TCP friendly" schemes, which can only achieve uniform bandwidth allocations; and 2) the majority of previously proposed flexible bandwidth-allocation schemes achieve arbitrary bandwidth allocations by prioritizing and scheduling packet transmissions within network equipment (i.e., within routers, base stations, etc.). In contrast, ABRC is an end-to-end application-layer solution that does not require changes to current WLAN products, making it more readily deployable over existing networks.

Abstract In this paper, a simple self‐adaptive contention window adjustment algorithm for 802.11 wireless local area networks (WLAN) is proposed and analyzed. Numerical results show that the new algorithm outperforms the standard 802.11 window adjustment algorithm. Compared with the standard and previously proposed enhancement algorithms, a salient feature of our algorithm is that it performs well in both heavy and light contention cases regardless of the packet sizes and physical versions. Moreover, the adaptive window adjustment algorithm is simpler than previously proposed schemes in that no live measurement of the WLAN traffic activity is needed. Copyright © 2004 John Wiley & Sons, Ltd.

Recently, a number of researchers have started to investigate new video-on-demand (VoD) architectures using batching, patching and periodic broadcasting. These architectures, compared to traditional unicast VoD systems, are much more scalable and can serve thousands or even millions of clients concurrently. Nevertheless, existing studies are usually focused on architectural issues. The problem of designing an efficient server to implement these new multicast VoD architectures has received little attention. While existing server designs using round-based schedulers can still be used, results show that such designs are suboptimal as they do not exploit the characteristics of fixed-schedule periodic broadcasting channels. We address this challenge by presenting an efficient server design for a recent multicast VoD architecture called Super-Scalar Video-on-Demand (SS-VoD). Results show that the efficient server design can increase the system capacity by 60% compared to traditional video server designs. We present details of this new server design, derives a performance model, and analyzes it using numerical results.

Recently, a new server-less architecture is proposed for building low-cost yet scalable video streaming systems. Compare to conventional client-server-based video streaming systems, this server-less architecture does not need any dedicated video server and yet is highly scalable. Video data are distributed among user hosts and these hosts cooperate to stream video data to one another. Thus as new hosts join the system, they also add streaming and storage capacity to absorb the added streaming load. This study investigates the data reorganization problem when growing a server-less video streaming system. Specifically, as video data are distributed among user hosts, these data will need to be redistributed to newly joined hosts to utilize their storage and streaming capacity. This study presents a new data reorganization algorithm that allows controllable tradeoff between data reorganization overhead and streaming load balance.

Video-on-demand (VoD) systems that can serve hundreds to thousands of concurrent users are already widely available. However, to deploy metropolitan-scale VoD services for potentially tens of millions of users, current VoD systems are still limited in capacity, and expensive in cost. To tackle this challenge, this study proposes a new periodic broadcasting scheme, called constrained consonant broadcasting (CCB), for large-scale video streaming. CCB outperforms all existing periodic broadcasting schemes while at the same time addresses two important constraints in practice, namely client access bandwidth and buffer requirements. For example, with a client access bandwidth constraint of twice the video bit-rate, a client buffer of 20% of the video size, and a total system bandwidth equal to six times the video bit-rate, the proposed CCB scheme can reduce the maximum startup latency by 72% and 70% compared to the current state-of-the-art skyscraper broadcasting and greedy disk-conserving broadcasting schemes respectively.

Recently, a new server-less architecture is proposed for building low-cost yet scalable video streaming systems. In this architecture, video blocks are distributed among user hosts and these hosts cooperate to stream video blocks to one another. To improve reliability, data and capacity redundancy are introduced to sustain node failures. However, the data placement as well as the redundant data in the system will need to be updated whenever new nodes join the system. Results show that the redundancy update overhead is very significant and even exceeds that in data reorganization. We present a new Transpositional Redundant Data Update algorithm that takes advantage of the structure of Reed-Solomon Erasure Correction codes and employs a special encoding scheme to significantly reduce the redundancy update overhead, especially when updates are performed in batch.

Video-on-demand (VoD) systems have traditionally been built around the client-server architecture, where a video server stores compressed video for delivery to clients connected by a network. As the system scales up, the server will need to be upgraded and this can become very expensive when the system scales beyond thousands of users. We investigate a radically different architecture where the bottleneck - video server, is eliminated altogether. Specifically, this server-less architecture relies on the client machines for distributed data storage and delivery. A client initiating a new streaming session will first locate other clients where the requested stream is stored, and then requests delivery of the stream directly from those clients instead of from a central server. This fully distributed architecture is inherently scalable as the storage and delivery capacity grows with the number of clients in the system.

In recent years, a number of sophisticated architectures have been proposed to provide VoD service using multicast transmissions. Compared to their unicast counterparts, these multicast VoD systems are highly scalable and can potentially serve millions of concurrent users. Nevertheless, these systems are designed for streaming constant-bit-rate (CBR) encoded videos and thus cannot benefit from the improved visual quality obtainable from variable-bit-rate (VBR) encoding techniques. To tackle this challenge, this paper presents a novel Slice-and-Patch (S&P) algorithm to support VBR video streaming in a multicast VoD system. Extensive trace-driven simulations are conducted to compare performance of the S&P algorithm with two other algorithms based on priority scheduling. Results show that the S&P algorithm outperforms the other two priority scheduling algorithms for most videos. Compared to the CBR counterpart serving videos of the same average bitrate, the S&P algorithm is able to support VBR video streaming with only 50% increase in latency. Given that VBR-encoded video can achieve visual quality comparable to CBR-encoded video at half the bitrate, this S&P algorithm can potentially achieve performance comparable to CBR-based systems when combined with VBR encoding techniques.

Video-on-demand (VoD) systems have traditionally been built on the client-server architecture, where a video server stores, retrieves, and transmits video data to video clients for playback This paper investigates a radically different approach to building VoD systems, one where the server, and hence the primary bottleneck, is completely eliminated. This server-less architecture comprises homogeneous hosts, called nodes, which serve both as client and as mini-server. Video data are distributed over all nodes and these nodes cooperatively stream video data to one another for playback. However, unlike traditional video server that runs on high-end server hardware in a carefully controlled and protected data centre, a node in a server less system is likely to be far more unreliable. Therefore it is essential that sufficient data and capacity redundancies are incorporated to maintain an acceptable set-vice reliability. This paper presents and analyzes a fault tolerant mechanism based on inter-node striping and erasure correction codes to tackle this challenge. By formulating the system's reliability as a Markov chain model, we obtain insights into the feasible operating region of the system, such as the amount of redundancy required and the node-level reliability that can be tolerated. Numerical results show that a server-less VoD system of 200 nodes can achieve reliability surpassing that of dedicated video server using a redundancy overhead of only 21.2% even though individual nodes are highly unreliable.

Voice communications such as telephony are delay sensitive. Existing voice-over-IP (VoIP) applications transmit voice data in packets of very small size to minimize packetization delay, causing very inefficient use of network bandwidth. This paper proposes a multiplexing scheme for improving the bandwidth efficiency of existing VoIP applications. By installing a multiplexer in an H.323 proxy, voice packets from multiple sources are combined into one IP packet for transmission. A demultiplexer at the receiver-end proxy restores the original voice packets before delivering them to the end-user applications. Results show that the multiplexing scheme can increase bandwidth efficiency by as much as 300%. The multiplexing scheme is fully compatible with existing H.323-compliant VoIP applications and can be readily deployed.

Unified video-on-demand (UVoD) is an architecture that integrates multicast transmission with unicast transmission to improve system efficiency. Streaming channels in a UVoD system are divided into unicast and multicast channels, with the multicast channels further divided equally among all videos. This uniform channel-allocation scheme is simple to design and implement, but the performance may not be optimal due to differences in video popularity. This paper investigates this channel-allocation problem with the goal of optimizing system efficiency. First, the uniform allocation assumption is removed and the channel-allocation problem formulated as a nonlinear integer optimization problem. This optimization model results in nonuniform channel allocations that can save up to 10% of channels. Second, to reduce the computational complexity in solving the nonlinear optimization model, an approximate model is derived and solved under small-latency conditions to obtain a closed-form solution. Third, a much simpler class-based popularity model is proposed and shown to achieve good efficiency, even if the precise popularity of each video is not known. Lastly, a zero-multicast channel-optimization algorithm is introduced that can further reduce channel requirement for systems with a large number of video selections. Numerical results show that optimized nonuniform channel-allocation policies can achieve channel reduction over uniform channel allocation by as much as 50% for a 1000-video system.

Despite the availability of video-on-demand (VoD) services in a number of cities around the world, large-scale deployment of VoD services in a metropolitan area is still economically impractical. This study presents a novel super-scalar architecture for building very large-scale and efficient VoD systems. The proposed architecture combines the use of static multicast, dynamic multicast, and intelligent client-side caching to vastly reduce server and network resource requirement. Moreover, in sharp contrast to conventional VoD systems where the system cost increases at least linearly with the system scale, the proposed architecture becomes more efficient as the system scales up and can ultimately be scaled up to serve any number of users while still keeping the startup latency short. This paper presents this new architecture, proposes methods to support interactive playback controls without the need for additional server or client resources, and derives an approximate performance model to relate the startup latency with other system parameters. The performance model is validated using simulation and the architecture is evaluated under various system settings. Lastly, a system implementation is presented and benchmarking results obtained to further verify the architecture, the performance model, and the simulation results.

Current video-on-demand (VoD)) systems can be classified into two categories: 1) true-Voll) (TVoD) and 2) near-VoD (NVod)). TVoD systems allocate a dedicated channel for every user to achieve short response times so that the user can select what video to play, when to play it, and perform interactive VCR-like controls at will. By contrast, NVoD systems transmit videos repeatedly over multiple broadcast or multicast channels to enable multiple users to share a single video channel so that system cost can be substantially reduced. The tradeoffs are limited video selections, fixed playback schedule, and limited or no interactive control. TVoD systems can be considered as one extreme where service quality is maximized, while NVoD systems can be considered as the other extreme where system cost is minimized. This paper proposes a novel architecture called Unified VoD) (UVoD) that can be configured to achieve cost-performance tradeoff anywhere between the two extremes (i.e., TVoD and NVoD). Assuming that a video client can concurrently receive two video channels and has local buffers for caching a portion of the video data, the proposed UVoD architecture can achieve significant performance gains (e.g., 400% more capacity for a 500-channel system) over TVoD under the same latency constraint. This paper presents the UVoD architecture, establishes a performance model, and analyzes UVoD's performance via numerical and simulation results.

With the rapid performance improvements in low-cost PCs, it becomes increasingly practical and cost-effective to implement large-scale video-on-demand (VoD) systems around parallel PC servers. This paper proposes a novel parallel video server architecture where video data are striped across an array of autonomous servers connected by an interconnection network. To coordinate data transmissions from multiple autonomous servers to a client station, a staggered push scheduling algorithm is proposed. A system model is constructed to quantify the performance of the architecture. Unlike most studies, this work does not assume the existence of a global clock among the servers and tackles two problems arising from server asynchrony: inconsistent schedule assignment and traffic overlapping. The former problem is solved by using an admission scheduler and the latter problem is solved by an over-rate transmission scheme. Analytical results prove a remarkable property of the staggered push architecture: as long as the network has sufficient capacity, the system can be scaled up linearly to an arbitrary number of servers. Design examples and numerical results are used to evaluate the proposed architecture under realistic assumptions and to compare it against other architecture.

Continuous-media (CM) servers have been around for some years. Apart from server capacity, another important issue in the deployment of CM servers is reliability. This study investigates rebuild algorithms for automatically rebuilding data stored in a failed disk into a spare disk. Specifically, a block-based rebuild algorithm is studied with the rebuild time and buffer requirement modeled. A buffer-sharing scheme is then proposed to eliminate the additional buffers needed by the rebuild process. To further improve rebuild performance, a track-based rebuild algorithm that rebuilds lost data in tracks is proposed and analyzed. Results show that track-based rebuild, while it substantially outperforms block-based rebuild, requires significantly more buffers (17-135 percent more) even with buffer sharing. To tackle this problem, a novel pipelined rebuild algorithm is proposed to take advantage of the sequential property of track retrievals to pipeline the reading and writing processes. This pipelined rebuild algorithm achieves the same rebuild performance as track-based rebuild, but reduces the extra buffer requirement to insignificant levels (0.7-1.9 percent). Numerical results computed using models of five commercial disk drives demonstrate that automatic rebuild of a failed disk can be done in a reasonable amount of time, even at relatively high server utilization (e.g., less than 1.5 hours at 90 percent utilization).

Parallel video servers have been proposed for building large-scale video-on-demand (VoD) systems from multiple low-cost servers. However, when adding more servers to scale up the capacity, system-level reliability will decrease as failure of any one of the servers will cripple the entire system. To tackle this reliability problem, this paper proposes and analyzes architectures to support server-level fault tolerance in parallel video servers. Based on the concurrent push architecture proposed earlier, this paper tackles three problems pertaining to fault tolerance, namely redundancy management, redundant data transmission protocol, and real-time fault masking. First, redundant data based on erasure codes are introduced to video data stored in the servers, which are then delivered to the clients to support fault tolerance. Despite the success of distributed redundancy striping schemes such as RAID-5 in disk array implementations, we discover that similar schemes extended to the server context do not scale well. Instead, we propose a redundant server scheme that is both scalable, and with lower total server buffer requirement. Second, two protocols are proposed to manage the transmission of redundant data to the clients, namely forward erasure correction which always transmits redundant data, and on-demand correction which transmits redundant data only after a server failure is detected. Third, to enable ongoing video sessions to maintain nonstop video playback during failure, we propose using fault masking at the client to recompute lost video data in real-time. In particular we derive the amount of client buffer required so that nonstop, continuous video playback can be maintained despite server failures.

There has been a trend toward designing video-on-demand systems using parallel-server architectures. By exploiting server-level parallelism, researchers can break through the performance limit of a single server, while keeping the system cost low by leveraging on commodity hardware platforms. A number of studies have demonstrated the feasibility of building parallel video servers around the client-pull architecture and one can even incorporate data redundancy into the system to sustain server-level failures. However, due to randomness of request arrivals and server processing time, dimensioning the server resource requirement is often difficult. This paper tackles the problem of buffer management and dimensioning for such a pull-based parallel video server. Using a generic buffer-pool model with worst-case analysis, upper bounds on the server buffer requirement are derived for a parallel-server design with multiple disks per server. The obtained bounds are independent of placement policy, video bit-rate, disk-scheduling discipline, and even number of servers in the system, making it applicable to a wide range of server designs. The analytical results also proved that the scalability of this pull-based server design will not be limited by the server buffer requirement.

In conventional video-on-demand systems, video data are stored in a video server for delivery to multiple receivers over a communications network. The video server's hardware limits the maximum storage capacity as well as the maximum number of video sessions that can concurrently be delivered. Clearly, these limits will eventually be exceeded by the growing need for better video quality and larger user population. This paper studies a parallel video server architecture that exploits server parallelism to achieve incremental scalability. First, unlike data partition and replication, the architecture employs data striping at the server level to achieve fine-grain load balancing across multiple servers. Second, a client-pull service model is employed to eliminate the need for interserver synchronization. Third, an admission-scheduling algorithm is proposed to further control the instantaneous load at each server so that linear scalability can be achieved. This paper analyzes the performance of the architecture by deriving bounds for server service delay, client buffer requirement, prefetch delay, and scheduling delay. These performance metrics and design tradeoffs are further evaluated using numerical examples. Our results show that the proposed parallel video server architecture can be linearly scaled up to more concurrent users simply by adding more servers and redistributing the video data among the servers.

Most existing commercial video servers are designed for a single server. Consequently, the capacity of the system in terms of maximum sustainable concurrent sessions is limited by the performance of the video server hardware. This paper proposes and analyzes the performance of a novel parallel video server architecture where video data are striped across an array of autonomous servers. The architecture allows one to build incrementally scalable video servers without video data replication. The proposed concurrent-push scheduling algorithm allows the system to integrate with quality of service guarantees provided by today's switching networks. In this paper, the striping policy, the service model, and the concurrent-push scheduling algorithm are presented. A system model is constructed to quantify three performance metrics, namely, server buffer requirement, client buffer requirement, and system response time. Results show that a simple extension of the server-push service model does not perform well under the parallel video server architecture. To improve system performance, a novel extension of the grouped sweeping scheme called the asynchronous grouped sweeping scheme (AGSS) is introduced. To further increase the scalability of the architecture, a new subschedule striping scheme (SSS) is introduced. With the proposed AGSS and SSS, our parallel video server architecture can be scaled up to more than 10000 concurrent users.

Previously, we considered the design and implementation of a server array system for delivering video-on-demand service on a computer network. Data blocks are striped across an array of autonomous servers, and each client station contacts the servers one by one to retrieve blocks for constructing its own video stream. The server array approach has the benefits of (1) scalable storage capacity and throughput, and (2) load sharing across the servers. In this paper, we study fault-tolerance issues in a server array system and propose the notion of redundant array of inexpensive servers (RAIS), which is a server-level counterpart to redundant array of inexpensive disks (RAID). While some concepts from RAID may be applicable, we show that RAIS faces new challenges as failure detection and recovery are handled by network protocols, subjecting to error, loss, non-deterministic delay, and bandwidth limitation. Through an implementation of RAIS for delivering multimedia world-wide-web services, we show that (1) server-level fault tolerance can be achieved, (2) continuous playback of video and audio can be maintained in spite of server failures, and (3) graceful service degradation can be implemented using our object striping scheme.

Most video-on-demand (VOD) systems use a single, powerful server to deliver video streams to users. We consider a novel server array approach for delivering video services on networks. Such an approach has the benefits of (l) greater system capacity, as individual server has individual disk and network channel, (2) scalability, as more clients can be supported by adding more servers without data duplication, and (3) fault tolerance, as server-level fault-tolerant and fault-recovery schemes can be devised. We describe our experiences and results in the implementation of a server-array-based VOD system. Our system now has four P5-90 servers serving 40 PC-486 stations using a 10 Mbps Ethernet switch. The system can deliver 40 simultaneous and independent 1.2 Mbps, 30 fps, full TV size, MPEG-1 video streams. Our results demonstrate that with careful protocol design, Ethernet can deliver continuous video and audio services.