Deep learning (DL) powered real-time applications usually need continuous training using data streams generated over time and across different geographical locations. Enabling data offloading among computation nodes through model training is promising to mitigate the problem that devices generating large datasets may have low computation capability. However, offloading can compromise model convergence and incur communication costs, which must be balanced with the long-term cost spent on computation and model synchronization. Therefore, this paper proposes EdgeC3, a novel framework that can optimize the frequency of model aggregation and dynamic offloading for continuously generated data streams, navigating the trade-off between long-term accuracy and cost. We first provide a new error bound to capture the impacts of data dynamics that are varying over time and heterogeneous across devices, as well as quantifying varied data heterogeneity between local models and the global one. Based on the bound, we design a two-timescale online optimization framework. We periodically learn the synchronization frequency to adapt with uncertain future offloading and network changes. In the finer timescale, we manage online offloading by extending Lyapunov optimization techniques to handle an unconventional setting, where our long-term global constraint can have abruptly changed aggregation frequencies that are decided in the longer timescale. Finally, we theoretically prove the convergence of EdgeC3 by integrating the coupled effects of our two-timescale decisions, and we demonstrate its advantage through extensive experiments performing distributed DL training for different domains.

Home anomaly monitoring is crucial for the elderly who live alone. A number of IoT-based home monitoring systems have been available, but most rely on privacy-intrusive cameras. With more and more concerns on privacy and security of human data, anomaly detection based on nonintrusive IoT devices becomes more desirable. Considering the elderly consumers, a low-cost system with good detection accuracy is further critical for the system's acceptability by elderly users. We propose a smart home monitoring system for living-alone senior citizens, relying on carefully designed, low-cost infrared sensor devices, as well as a cloud-based data processing and anomaly detection platform. Our PIR sensor device is effective in continuous monitoring of motion data in a user's apartment, and an open-hardware software platform is devised to support sensors manufactured by various vendors in the IoT system, all for cost reduction purpose. For privacy preservation, we encrypt collected data and store data indices in a blockchain system, to achieve efficient data access control and auditing. For motion anomaly detection, we propose a simple but effective environment adaptation method to work with the one-class support vector machine (OCSVM) method. Experiments driven by real-world traces show good reliability, accuracy, and efficiency of our system.

In this paper, we investigate distributed broadcasting in dynamic networks, where the topology changes continually over time. We propose a network model that captures the dynamicity caused by both churn and mobility of nodes. In contrast to existing work on dynamic networks, our model defines the dynamicity in terms of localized topological changes in the vicinity of each node, rather than a global view of the whole network. Obviously, a local dynamic model suits distributed algorithms better than a global one. The proposed dynamic model uses the more realistic SINR model to depict wireless interference, instead of oversimplified graph-based models adopted in most existing work. We consider the fundamental communication primitive of global broadcast, which is to disseminate a message from a source node to the whole network. Specifically, we present a randomized distributed algorithm that can accomplish dynamic broadcasting in an asymptotically optimal running time of$O(D_{T})$with a high probability guarantee, under the assumption of reasonably constantdynamicity rate, where$D_{T}$is thedynamic diameter, a parameter proposed to depict the complexity of dynamic broadcasting. We believe our local dynamic model can greatly facilitate distributed algorithm studies in mobile and dynamic wireless networks.

We study the fundamental problem of fault-tolerant distributed sampling towards uniform probabilistic distribution in dynamic multi-hop wireless networks. Whereas uniform sampling has been extensively studied without concerning fault tolerance, only quite few proposals investigate how the uniform sampling algorithm tolerate Byzantine faults on dynamic networks with very special topologies, e.g., regular graphs with constant node degree. Therefore, designing fault-tolerant uniform sampling algorithms for more general graphs is still an open problem. To this end, we propose a fast and highly fault-tolerate randomized algorithm, such that nearly-uniform sampling is achieved in O(log2n) rounds, while up to O(√n/(polylog(n)·Δ)) Byzantine nodes can be tolerated, where Δ is the maximum degree of the network. Moreover, the proposed algorithm is also communication efficient in the sense that only O(log n) bits need to be exchanged on each link in every round. To show the power of distributed uniform sampling, we apply the proposed algorithm in designing polylogarithmic time distributed algorithms for two typical fundamental issues, i.e., to achieve agreement or data aggregation in Byzantine dynamic networks.

This paper studies jamming-resilient distributed backbone construction in multi-hop wireless networks. Specifically, a strong adversarial jamming model is proposed that captures the general jamming phenomena suffered by wireless communications. The jamming model is based on the realistic Signal-to-Interference-plus-Noise-Ratio (SINR) interference model, and is featured by local-uniformity, unrestricted energy budget and reactivity, which covers more jamming scenarios and is much closer to reality than existing jamming models. Under the strong adversarial jamming model, we propose a randomized distributed algorithm that can construct a backbone in J(O(log n + logR)) rounds with high probability, where J(O(log n + log R)) is the number of rounds in the interval from the beginning of the algorithm execution that contains O(log n + log R) unjammed rounds for every node. This result is asymptotically optimal considering the trivial lower bound of Ω(log n) for a successful transmission even without interference and jamming.

This paper investigates distributed Dominating Set (DS) and Connected Dominating Set (CDS) construction in dynamic wireless networks under the SINR interference model. Specifically, we present a new model for dynamic networks that admits both churns (due to node arrivals/departures) and node mobility. Under this dynamic model, we propose efficient algorithms to construct a DS and a CDS with constant approximation ratios w.r.t. the corresponding minimum ones in O(log n) time with a high probability guarantee. To the best of our knowledge, these algorithms are the first known ones for DS and CDS construction in dynamic networks assuming the SINR interference model. We believe our dynamic network model can greatly facilitate distributed algorithm studies in mobile and dynamic wireless networks.

Dynamic Adaptive Streaming over HTTP (DASH) has emerged as a popular approach for video transmission, which brings a potential benefit for the Quality of Experience (QoE) because of its segment-based flexibility. However, the Internet can only provide no guaranteed delivery. The high dynamic of the available bandwidth may cause bitrate switching or video rebuffering, thus inevitably damaging the QoE. Besides, the frequently requested popular videos are transmitted for multiple times and contribute to most of the bandwidth consumption, which causes massive transmission redundancy. Therefore, we propose a Learning-based Edge with cAching and Prefetching (LEAP) to improve the online user QoE of adaptive video streaming. LEAP introduces caching into the edge to reduce the redundant video transmission and employs prefetching to fight against network jitters. Taking the state information of users into account, LEAP intelligently makes the most beneficial decisions of caching and prefetching by a QoE-oriented deep neural network model. To demonstrate the performance of our scheme, we deploy the implemented prototype of LEAP in both the simulated scenario and the real Internet. Compared with all selected schemes, LEAP at least raises average bitrate by 34.4% and reduces video rebuffering by 42.7%, which leads to at least 15.9% improvement in the user QoE in the simulated scenario. The results in the real Internet scenario further confirm the superiority of LEAP.

As it has become the norm for cloud providers to host multiple datacenters around the globe, significant demands exist for inter-datacenter data transfers in large volumes, e.g., migration of big data. A challenge arises on how to schedule the bulk data transfers at different urgency levels, in order to fully utilize the available inter-datacenter bandwidth. The Software Defined Networking (SDN) paradigm has emerged recently which decouples the control plane from the data paths, enabling potential global optimization of data routing in a network. This paper aims to design a dynamic, highly efficient bulk data transfer service in a geo-distributed datacenter system, and engineer its design and solution algorithms closely within an SDN architecture. We model data transfer demands as delay tolerant migration requests with different finishing deadlines. Thanks to the flexibility provided by SDN, we enable dynamic, optimal routing of distinct chunks within each bulk data transfer (instead of treating each transfer as an infinite flow), which can be temporarily stored at intermediate datacenters to mitigate bandwidth contention with more urgent transfers. An optimal chunk routing optimization model is formulated to solve for the best chunk transfer schedules over time. To derive the optimal schedules in an online fashion, three algorithms are discussed, namely a bandwidth-reserving algorithm, a dynamically-adjusting algorithm, and a future-demand-friendly algorithm, targeting at different levels of optimality and scalability. We build an SDN system based on the Beacon platform and OpenFlow APIs, and carefully engineer our bulk data transfer algorithms in the system. Extensive real-world experiments are carried out to compare the three algorithms as well as those from the existing literature, in terms of routing optimality, computational delay and overhead.

The hardware improvement of mobile devices and pervasiveness of wireless technology expedite the convergence with the fast growing cloud computing trend, where the abundant resources on the cloud meet well with the deficiency of hand-held devices. Cloudlet, as a newly emerging paradigm “bringing the cloud closer” to the end users, features a more scalable deployment fashion where idle personal servers can be efficiently harnessed. Despite an envisioned monetary saving, such a paradigm confines itself to limited application scenarios, which fails to reach a wide realm of roaming users outside the distance coverage of the access points. In this paper, we propose a cloudlet-based multi-lateral resource exchange framework for mobile users, relying on no central entities. Inspired by the success of BitCoin, we design a novel virtual currency tailored for our framework. To realize an efficient resource exchange market, we also introduce flexible pricing strategies adopted by the individual users whom we assume are rational price-takers, with solid theoretical analysis on the equilibrium state and the stability. After elaborating the key functional modules, we introduce a prototype design enabling seamless trading among mobile users on Internet bandwidth as a proof-of-concept, with least user intervention. Both simulations and experiments are conducted to verify the practicality and efficiency of our system.

Federation of geo-distributed cloud services is a trend in cloud computing that, by spanning multiple data centers at different geographical locations, can provide a cloud platform with much larger capacities. Such a geo-distributed cloud is ideal for supporting large-scale social media applications with dynamic contents and demands. Although promising, its realization presents challenges on how to efficiently store and migrate contents among different cloud sites and how to distribute user requests to the appropriate sites for timely responses at modest costs. These challenges escalate when we consider the persistently increasing contents and volatile user behaviors in a social media application. By exploiting social influences among users, this paper proposes efficient proactive algorithms for dynamic, optimal scaling of a social media application in a geo-distributed cloud. Our key contribution is an online content migration and request distribution algorithm with the following features: 1) future demand prediction by novelly characterizing social influences among the users in a simple but effective epidemic model; 2) one-shot optimal content migration and request distribution based on efficient optimization algorithms to address the predicted demand; and 3) a Δ(t)-step look-ahead mechanism to adjust the one-shot optimization results toward the offline optimum. We verify the effectiveness of our online algorithm by solid theoretical analysis, as well as thorough comparisons to ready algorithms including the ideal offline optimum, using large-scale experiments with dynamic realistic settings on Amazon Elastic Compute Cloud (EC2).

The rapidly increasing power of personal mobile devices (smartphones, tablets, etc.) is providing much richer contents and social interactions to users on the move. This trend however is throttled by the limited battery lifetime of mobile devices and unstable wireless connectivity, making the highest possible quality of service experienced by mobile users not feasible. The recent cloud computing technology, with its rich resources to compensate for the limitations of mobile devices and connections, can potentially provide an ideal platform to support the desired mobile services. Tough challenges arise on how to effectively exploit cloud resources to facilitate mobile services, especially those with stringent interaction delay requirements. In this paper, we propose the design of a Cloud-based, novel Mobile sOcial tV system (CloudMoV). The system effectively utilizes both PaaS (Platform-as-a-Service) and IaaS (Infrastructure-as-a-Service) cloud services to offer the living-room experience of video watching to a group of disparate mobile users who can interact socially while sharing the video. To guarantee good streaming quality as experienced by the mobile users with time-varying wireless connectivity, we employ a surrogate for each user in the IaaS cloud for video downloading and social exchanges on behalf of the user. The surrogate performs efficient stream transcoding that matches the current connectivity quality of the mobile user. Given the battery life as a key performance bottleneck, we advocate the use of burst transmission from the surrogates to the mobile users, and carefully decide the burst size which can lead to high energy efficiency and streaming quality. Social interactions among the users, in terms of spontaneous textual exchanges, are effectively achieved by efficient designs of data storage with BigTable and dynamic handling of large volumes of concurrent messages in a typical PaaS cloud. These various designs for flexible transcoding capabilities, battery efficiency of mobile devices and spontaneous social interactivity together provide an ideal platform for mobile social TV services. We have implemented CloudMoV on Amazon EC2 and Google App Engine and verified its superior performance based on real-world experiments.

Federation of geo-distributed cloud services is a trend in cloud computing which, by spanning multiple data centers at different geographical locations, can provide a cloud platform with much larger capacities. Such a geo-distributed cloud is ideal for supporting large-scale social media streaming applications (e.g., YouTube-like sites) with dynamic contents and demands, owing to its abundant on-demand storage/bandwidth capacities and geographical proximity to different groups of users. Although promising, its realization presents challenges on how to efficiently store and migrate contents among different cloud sites (i.e. data centers), and to distribute user requests to the appropriate sites for timely responses at modest costs. These challenges escalate when we consider the persistently increasing contents and volatile user behaviors in a social media application. By exploiting social influences among users, this paper proposes efficient proactive algorithms for dynamic, optimal scaling of a social media application in a geo-distributed cloud. Our key contribution is an online content migration and request distribution algorithm with the following features: (1) future demand prediction by novelly characterizing social influences among the users in a simple but effective epidemic model; (2) oneshot optimal content migration and request distribution based on efficient optimization algorithms to address the predicted demand, and (3) a Δ(t)-step look-ahead mechanism to adjust the one-shot optimization results towards the offline optimum. We verify the effectiveness of our algorithm using solid theoretical analysis, as well as large-scale experiments under dynamic realistic settings on a home-built cloud platform.

Internet-based cloud computing is a new computing paradigm aiming to provide agile and scalable resource access in a utility-like fashion. Other than being an ideal platform for computation-intensive tasks, clouds are believed to be also suitable to support large-scale applications with periods of flash crowds by providing elastic amounts of bandwidth and other resources on the fly. The fundamental question is how to configure the cloud utility to meet the highly dynamic demands of such applications at a modest cost. In this paper, we address this practical issue with solid theoretical analysis and efficient algorithm design using Video on Demand (VoD) as the example application. Having intensive bandwidth and storage demands in real time, VoD applications are purportedly ideal candidates to be supported on a cloud platform, where the on-demand resource supply of the cloud meets the dynamic demands of the VoD applications. We introduce a queueing network based model to characterize the viewing behaviors of users in a multichannel VoD application, and derive the server capacities needed to support smooth playback in the channels for two popular streaming models: client-server and P2P. We then propose a dynamic cloud resource provisioning algorithm which, using the derived capacities and instantaneous network statistics as inputs, can effectively support VoD streaming with low cloud utilization cost. Our analysis and algorithm design are verified and extensively evaluated using large-scale experiments under dynamic realistic settings on a home-built cloud platform.