Real-time video analytics (VA) presents challenges due to its computational intensity and latency sensitivity, especially when processed on mobile devices with limited local resources. We propose to offload VA tasks to edge servers with diverse computational capabilities. We present a "detect + track" approach with on-device object tracking and edge-assisted object detection. We formulate a long-term nonlinear integer programming to maximize the overall accuracy within detection frequency and latency constraints. We then design a queue-based online optimization algorithm to solve it: relax the original problem from the integer domain to the real domain, then employ a queue-based adaptation and randomized rounding strategy. Via rigorous proof, both dynamic regret regarding detection accuracy and the real-time requirement are ensured. Evaluation results also demonstrate the effectiveness of our approach.

The demand for video analytics is increasing rapidly. Due to the limited computational and network resources on edge servers, adjusting video configurations such as resolution and frame rate has become an effective strategy to reduce computational and transmission costs. However, this can also compromise detection accuracy, necessitating a balance between resource consumption and analytics accuracy. Also, the dynamic availability of edge servers and variability in their energy consumption further complicates making offloading decisions and configuration selection. In this paper, we first model the problem as a mixed planning program. Then we propose a volatile MAB-based configuration selection algorithm, VACS, which aims to maximize video analytics accuracy while reducing the overall energy consumption. Rigorous proof measures the gap between online decisions and the optimum. Extensive experiments validate the effectiveness of VACS.

Video analytics pipelines migrating to edge deployments are facing performance bottlenecks under limited bandwidth. Non-uniform intra-frame encoding emerges to further compress pixels without affecting the output of the server deep neural network (DNN), while it is inefficient in high-resolution video streaming at low bandwidth. The detail enhancement capability of neural super-resolution (SR) permits resolution downsampling and aggressive compression on edge devices for low-latency transmission. To exploit its accuracy potential, DNN-oriented non-uniform encoding is expected to be additionally aware of SR models. However, traditional codecs struggle to cope with both quality optimization for SR and global semantic features for DNN. We advocate neural codecs for coordinated encoding and enhancement, enabling analytic-oriented video streaming with optimal accuracy-delay tradeoffs. Our system, VidIQ, achieves quality-enhanced real-time video analytics by 1) improving the network architecture of neural codecs (at two granularity) to integrate SR models into a DNN-oriented analytics pipeline, and 2) adapting the multi-scale encoder and SR-decoder to scene dynamics (i.e., content and bandwidth variations) with the help of the monolithic controller to hold a performance advantage. Extensive evaluations showcase that VidIQ reduces end-to-end delay by 35.8% and improves analytics accuracy by 21.2% compared to the recent video compression, enhancement, and streaming baselines.

Video analytics pipelines (VAPs) have been a paradigm for large-scale video analytics. Due to temporal redundancy in video, frame filtering is widely used in VAPs to reduce analysis workload. However, existing works overlook a limitation: while inference operates only on selected frames, decoders must still process many redundant frames due to codec dependencies, leading to over-decoding trap. This limitation stems from the reference-based design in modern codecs, which require decoding preceding frames to reconstruct any selected one. As a result, over-decoding has become the practical bottleneck in VAPs using modern decoders, highlighting a critical but under-explored problem. To address this issue, we propose ParaDeco, a high-throughput video analytics framework featuring a novel frame-level parallel generative decoder. Unlike traditional decoders, ParaDeco adopts a decode-what-matters approach with decoupled frame dependencies. To decode arbitrary frames independently, ParaDeco generates frame-wise features as standalone skeletons using compressed video metadata, then predicts pseudo frames maintaining semantic consistency with original frames. Moreover, ParaDeco identifies which frames truly matter for analysis via delicate contribution-based frame filtering. We implement ParaDeco on a cloud server and evaluate it on large-scale real-world video datasets. Our experimental results show that ParaDeco achieves a 2.76× speedup on average compared to state-of-the-art VAPs.

We are now embracing an era of edge computing and artificial intelligence, and the combination of the two has spawned a new field of research called edge intelligence. Massive amounts of data is generated at the edge of network, which relies on artificial intelligence to realize its potential. Meanwhile, artificial intelligence is able to flourish when processing diverse edge data. However, the computation and storage resources of edge servers are not unlimited. For some large-scale intelligent applications, it is difficult to meet their service quality requirements by directly offloading the entire application to a nearby server for processing. Due to the heterogeneity of server resources in edge environments, how to balance the workload among edge servers to provide better services also becomes complicated. The goal of this paper is to minimize the total cost of offloading large-scale applications consisting of many dependent tasks in an edge system. We formulate the Dependent task Offloading with Service Caching (DOSC) problem, which is proved to be NP-hard. A dynamic planning-based algorithm is introduced to solve fixed-DOSC, in which some services are pre-configured on the edge server, and other services can not be downloaded from the remote cloud. We also present a theoretical analysis on the performance guarantee of the dynamic planning-based algorithm. Then, we propose a near-optimal algorithm using the Gibbs sampling to solve the general DOSC problem. Testbed experiments and trace-driven simulations are conducted to verify the performance of our algorithm. Our algorithm, shown to be the most effective in terms of cost, considers both service caching and task dependencies when task offloading in comparison to other baseline algorithms.

Massively deployed cameras form a tightly connected network which generates video streams continuously. Benefiting from advances in computer vision, automated real-time analytics of video streams can be of practical value in various scenarios. As cameras become more dense, cross-camera video analytics has emerged. Combining video contents from multiple cameras for analytics is certainly more promising than single-camera analytics, which can realize cross-camera pedestrian tracking and cross-camera complex behavior recognition. Some works focused on optimization of cross-camera video analytic applications, but most of them ignore specific network situation between cameras and edge servers. Furthermore, most of them ignore the super resolution technique, which is proven to be a source of efficiency. In this paper, we first verify the potential gain of super resolution on cross-camera video analytic tasks. Then, we design and implement a cross-camera real-time video streaming analytic system,${\mathsf {Scrava}}$, which leverages super resolution to augment low-resolution videos and simultaneously reduce bandwidth consumption.${\mathsf {Scrava}}$enables real-time cross-camera video analytics and enhances video segments with the SR module under poor network conditions. We take cross-camera pedestrian tracking as an example, and experimentally verifies the effectiveness of super resolution on real-time cross-camera video analytics. Compared with using low-resolution video segments,${\mathsf {Scrava}}$can improve the F1 score by 47.16%, verifying the feasibility of exploiting super resolution to improve the performance of real-time cross-camera video analytic systems.

Edge video analytics typically rely on conventional encoding standards to transmit device visual data for server-side inference. Unfortunately, general-purpose compression solutions retain unnecessary visual data that does not contribute to accuracy, resulting in significant latency throughout Video Analytics Pipeline (VAP). While previous approaches have made partial progress, they cannot systematically eliminate VAP redundancy due to uncoordinated subsystem-level optimization. Achieving complete redundancy elimination presents a major challenge, as a lack of spatio-temporal coordination risks offsetting latency gains with computational overhead (associated with redundancy elimination).Crucioovercomes these limitations with an end-to-edge framework that integrates temporally adaptive frame filtering and coordinated video compression. It leverages redesigned asymmetric autoencoders to synchronize inter-frame temporal compression with intra-frame spatial feature extraction. Additionally,Crucioemploys a one-pass decoding mechanism for encoded critical frames and dynamically adjusts batching scales to minimize latency. Empirical results demonstrateCrucio's superiority, outperforming existing solutions (e.g., DDS, Reducto, and STAC) by over a 31% reduction in end-to-end latency at 0.9 accuracy thresholds.

In recent years, many crowdsourcing platforms have emerged, using the resources of recruited workers to perform diverse outsourcing tasks, where the video analytics attracts much attention due to its practical implications. For maximum profits, platforms carefully choose the workers and determine the video analytics configurations to ensure accuracy; meanwhile, workers possess the flexibility to tailor the configurations for their indivi-dual gains, which makes it hard for platforms to optimize their profits considering the platform-worker conflicts. In this paper, we design an incentive mechanism for Multi-leader game-based video Analytics upon CROwdsourcing, named MACRO, to over-come the above situation. Under that mechanism, we first formu-late the utility optimization problems for platforms and workers, respectively. We then propose a dual ascent-based method to op-timally determine the video analytics configurations for a multi-platform game, ensuring Pareto efficiency. Moreover, in the context of a multi-leader game involving platform-worker conflicts, we design an incentive function with its incentive factor update strategy and propose an ADMM-based approach for maximizing incentives that motivate workers to contribute to the platforms' profits. Rigorous proofs demonstrate the linear convergence of the MACRO to the multi-leader Stackelberg equilibrium. Trace-driven experiments show that MACRO improves the Pareto efficiency by 26.3%, outperforming other approaches.

Recent years have witnessed the unprecedented performance of convolutional networks in image super-resolution (SR). SR involves upscaling a single low-resolution image to meet application-specific image quality demands, making it vital for mobile devices. However, the excessive computational and memory requirements of SR tasks pose a challenge in mapping SR networks on a single resource-constrained mobile device, especially for an ultra-high target resolution. This work presents TileSR, a novel framework for efficient image SR through tile-granular parallel offloading upon multiple collaborative mobile devices. In particular, for an incoming image, TileSR first uniformly divides it into multiple tiles and selects the top-K tiles with the highest upscaling difficulty (quantified by mPV). Then, we propose a tile scheduling algorithm based on multi-agent multiarmed bandit, which attains the accurate offload reward through the exploration phase, derives the tile packing decision based on the reward estimates, and exploits this decision to schedule the selected tiles. We have implemented TileSR fully based on COTS hardware, and the experimental results demonstrate that TileSR reduces the response latency by 17.77-82.2% while improving the image quality by 2.38-10.57% compared to other alternatives.

There is an increasing demand for analyzing live surveillance video streams via large-scale camera networks, particularly for applications in public safety and smart cities. To address the conflict between resource-intensive detection models and limited capabilities of cameras, a detection-with-tracking framework has gained prominence. However, since trackers are vulnerable to occlusions and new object appearances, frequent detections are required to calibrate the results, leading to varying detection demands that depends on video content. Consequently, we propose a mechanism for content-aware analytics on collaborative cameras, denoted as VisFlow, to increase the quality of detections and achieve the latency requirement by fully utilizing camera resources. We formulate such a problem as a non-linear, integer program with a long-term perspective, aimed at maximizing detection accuracy. An online mechanism, underpinned by a queue-based algorithm and randomized rounding, is then devised to dynamically orchestrate detection workloads among cameras, thus adapting to fluctuating detection demands. Via rigorous proof, both dynamic regret regarding overall accuracy and the transmission budget are ensured in the long run. The testbed experiments on Jetson Kits demonstrate that VisFlow improves accuracy by 18.3% over the baselines.

Over the last decade, the mobile crowdsourcing has become a paradigm to conduct the manual annotation and further analytics by recruited workers, with their rewards depending on the result quality. Existing dispatchers cannot precisely capture the resource-quality trade-off for video analytics, because the configurations supported by recruited workers are limited, and workers’ availability changes over time. To determine the most suitable configurations as well as workers for video analytics, we formulate a non-linear mixed program in long term, maximizing the crowdsourcing profit. Based on previous results under various configurations and workers, we design an algorithm via a series of subproblems to decide the configurations adaptively upon the prediction of workers’ feedbacks. Such prediction is based on volatile multi-armed bandit to capture workers’ availability and stochastic changes on resource uses. Furthermore, we extend the proposed algorithms to the multi-worker selection scenario where the platform needs to determine a candidate worker set instead of a single worker for video analytics. Via rigorous proof, the regret is ensured upon the Lyapunov optimization and the bandit, measuring the gap between the online decisions and the offline optimum. Extensive trace-driven experiments show that our proposed algorithm improves the profit by 37% compared with other algorithms.

Deep convolutional neural network (NN)-based object detectors are not appropriate for straightforward inference on high-resolution videos at edge devices, as maintaining high accuracy often brings about prohibitively long latency. Although existing solutions have attempted to reduce on-device inference latency by selecting a cheaper configuration (e.g., choosing a more lightweight NN or scaling a frame to a smaller size before inference) or eliminating a background containing no object, they often ignore various high-resolution features and fail to optimize for those videos. We thus present AdaPyramid, a framework to reduce as much on-device inference latency as possible, especially for high-resolution videos, while achieving the accuracy demand approximately. We observe that the cheapest configuration to achieve the accuracy demand varies significantly across both different frames and different regions in a frame. The underlying reason is that object features (e.g., the location, size and category of objects) are more uneven in high-resolution videos, both temporally and spatially. Moreover, we observe that the object size presents a prominent hierarchical distribution in high-resolution frames. AdaPyramid thus partitions each frame hierarchically just like a pyramid and chooses a content-aware configuration for each region, which is adapted online based on the feedback. We evaluate the performance of AdaPyramid on a public dataset and our collected real-world videos. The obtained results show that under comparable accuracy to the state-of-the-art solutions, AdaPyramid can decrease inference latency by 40% on average, with up to 2.5× speed-up.

The usage of live streaming services has led to a substantial increase in live video traffic. However, the perceived quality of experience of users is frequently limited by variations in the upstream bandwidth of streamers. To address this issue, several adaptive bitrate (ABR) algorithms have been developed to mitigate bandwidth variations. Nevertheless, the ability of users to enjoy high-quality live streams remains limited. While neural-enhanced approaches, such as super-resolution, offer significant quality improvements, frame-oriented super-resolution leads to excessive inference delay that violates the real-time feature of live streaming. In response, we propose ViChaser, which examines block-oriented super-resolution for live streaming. ViChaser performs neural super-resolution on potential blocks of interest in the media server, corresponding to the user’s viewpoint, and uses online learning to adapt to the dynamic content of the video. Additionally, ViChaser utilizes the Lyapunov framework to efficiently allocate uplink bandwidth for original low-quality live video and high-quality labels. The experimental results demonstrate that ViChaser achieves 1.2–1.5 dB higher video quality in Peak-Signal-to-Noise-Ratio than WebRTC and increases processing speed by 11–16 fps relative to LiveNAS.

The widespread use of 5G and artificial intelligence applications has led to strong momentum in Mobile Edge Computing (MEC). With MEC, we can offload compute-intensive tasks to edge servers that are closer to the user, thereby reducing the long latency incurred by data transmission via WAN. Although many works have investigated task offloading decisions under service caching, the state of services is an equal, if not more important, research area of MEC, yet receive much less attention. In general, the arrival of tasks exhibit a distribution over time. Besides the necessary energy consumption in processing tasks offloaded to edge servers, a large amount of energy is required for maintaining services cached on servers. When more and more services become idle, they will incur a non-negligible additional energy. In this paper, we focus on an interesting but currently less studied problem in MEC, namely online service caching and state management in MEC. We propose DCSO, a bounded online algorithm that considers dynamic service caching and state management of services to minimize long-term cost in MEC systems. Meanwhile, our algorithm achieves a 2 competitive ratio in state management. Trace-driven simulations show that our algorithm reduces the overall cost efficiently while keeping low computation latency.

To address the emerging security challenges in Multi-Access Edge Computing (MEC), it is imperative that solutions go beyond the current infrastructure-centric measures. These methods, including authentication and access control, are insufficient to combat malware that conceals itself within ME applications. The acknowledged flaws in the ME application layer necessitate an immediate call for creative solutions. In this work, we propose a non-invasive security architecture for MEC, meticulously designed to strike a balance between performance burden and security protection capabilities. The objective of the design contains three major aspects, i.e. user experience, service density and serviceability. We conduct a thorough evaluation that enables us to quantify the significance of high bandwidth, low user experience latency and MEC serviceability. The experimental results and ablation studies indicate that our proposed method effectively balances user experience and security capabilities. This not only provides a practical and cost-effective solution but also establishes a strong precedent for the community to develop a secure MEC with superior performance in real-world production environments.

Deploying microservice instances on the edge device close to end users can provide on-site processing thus reducing request response time. Each microservice has multiple instances that can process requests in parallel. To achieve high processing efficiency, the number of these instances is scaled according to the workload, which is also known as autoscaling. Previous studies of microservice autoscaling in the edge computing environment lack in-depth consideration of time-varying workload, they assume that the workload of each microservice always depends on that of its upstream. However, through an analysis of Alibaba's microservice trace with hundreds of millions of records, we find that the assumption is impractical thus hurting autoscaling effectiveness. To solve this problem, we propose ProScale, a prediction-driven proactive autoscaling framework for microservices at the edge. ProScale proactively forecasts the workload for each individual microservice per timeslot. Then it utilizes an efficient online algorithm to leverage the predicting results to determine the instance number for each microservice jointly with making placement decisions. For each microservice instance deployed on the edge device, ProScale handles burst requests using a designed offloading strategy. In addition, ProScale can also balance the load for multiple instances of each microservice. Extensive trace-driven experiments show that ProScale has great scalability. It can reduce average response time by 96.7% and resource usage by 96.5% compared with existing strategies and designed baselines.

Distributed Interactive Applications (DIAs) such as virtual reality and multiplayer online game usually require fast processing of tremendous data and timely exchange of delay-sensitive action data and metadata. This makes traditional mobile-based or cloud-based solutions no longer effective. Thanks to edge computing, DIA Service Providers (DSPs) can rent resources from Edge Infrastructure Providers (EIPs) to place service entities that store user states and run computation-intensive tasks. One fundamental problem for a DSP is to decide where to place service entities to achieve low-delay pairwise interactions between DIA users, under the constraint that the total placement cost is no more than a specified budget threshold. In this article, we formally model the service entity placement problem and prove that it is NP-complete by a polynomial reduction from the set cover problem. We present GPA, an efficient algorithm for service entity placement, and theoretically analyze its performance. We evaluated GPA with both real-world data trace-driven simulations, and observed that GPA performs close to the optimal algorithm and generally outperforms the baseline algorithm. We also output a curve showing the trade-off between the weighted average interaction delay and the budget threshold, so that a DSP can choose the right balance.

In this paper, we consider a fundamental problem: given one mobile charger that can charge multiple sensor nodes simultaneously, how we can schedule it to charge a given WSN to maximize the energy usage effectiveness (EUE)? We propose a novel charging paradigm-Overlapped Mobile Charging (OMC)- the first of its kind to the best of our knowledge. Firstly, OMC clusters sensor nodes into multiple non-overlapped sets using k-means evaluated by the Davies-Bouldin Index, such that the sensor nodes in each set have similar recharging cycles. Secondly, for each set of sensor nodes, OMC further divides them into multiple overlapped groups, and charges each group at different locations for different time durations to make sure that each overlapped sensor node just receives its required energy from multiple charging locations.

Edge service entity placement is a fundamental issue in mobile edge computing, which tries to place service entities on edge servers to achieve better economic benefits and quality of service for users. Most existing studies towards this issue usually deploy application services separately; however, we observe that many application services can be broken down into smaller service components/entities, which may enable us to share these smaller entities between application services. Therefore, in this paper, we propose the concept of service entity chain, which is a chain of ordered service entities that represent an application service. We study the problem of placing service entities in the form of chains on edge servers within a given cost budget, so as to minimize the total latency experienced by users. We provide a formal problem formulation and design an efficient algorithm for it. Extensive simulations are conducted to demonstrate the advantages of the proposed algorithm compared with two state-of-the-art algorithms.

Virtualization techniques help edge environments separate the role of the traditional edge providers into two: edge infrastructure providers (EIPs), who manage the physical edge infrastructure, and edge service providers (ESPs), who aggregate resources (especially, compute resources) from multiple EIPs to place service entities and offer value-added services to end users (EUs). In such an environment, end users submit their data analysis jobs to ESPs; ESPs process the data analysis jobs using their service entities. One fundamental and critical problem for an ESP is to decide how much compute resources to rent from each edge server under the constraint that the total amount of rental resources is no more than a specified budget threshold, so that the average makespan of the data analysis jobs submitted to it is minimized. This Edge Resource Allocation (ERA) problem is proven to be NP-complete by reducing the set cover problem to a special case of it. To design an approximation algorithm for ERA, we perform two transformations on ERA: first, we transform ERA into mERA by replacing minimization with maximization; second, we transform mERA into dmERA by limiting the possible amounts of rental resources to a finite set of values. We find that dmERA has several tractable properties that allow us to design Hermes, a provably efficient algorithm that approximates the optimal allocation. We demonstrate that the gap between Hermes and the optimum in simulations and Android-based testbed experiments are no larger than 4.78% and 12.43%, respectively. Hermes can also output a curve showing the trade-off between the average makespan and the budget threshold, so that an ESP can choose the right balance.

Edge computing is one of the emerging technologies aiming to enable timely computation at the network edge. With virtualization technologies, the role of the traditional edge providers is separated into two: edge infrastructure providers (EIPs), who manage the physical edge infrastructure, and edge service providers (ESPs), who purchase slices of physical resources (e.g., CPU, bandwidth, memory space, disk storage) from EIPs and then cache service entities to offer their own value-added services to end users. When an ESP caches a service entity in an edge server, the ESP has to pay some fees (i.e, the cache cost) to the EIP that owns the edge server. One of the fundamental problems in edge virtualization is the so-called service entity caching problem, i.e., where to place service entities for an ESP to minimize the cache cost. In this paper, we study the service entity caching problem from the utility perspective. We use `utility' to denote the positive impact on a client from caching a service entity in an edge server, and the exact meaning of utility can vary depending on specific scenarios. We formulate the Utility-based Service Entity Caching (UtilitySEC) problem, which can be generalized to many existing problems by modifying the `utility'. We prove that the UtilitySEC problem is NP-complete and design an approximation algorithm for it. Extensive simulations are conducted to evaluate the performance of the proposed framework.

The lines between IaaS, PaaS, and SaaS are becoming blurred as datacenter providers seek to create cloud platforms that can widen their appeal to developers. With this kind of hybrid datacenter, resource requests from tenants are increasingly transforming into hybrid requests that may simultaneously demand IaaS, PaaS, and SaaS resources. This paper tackles the challenge of modeling and deploying hybrid tenant requests in datacenter networks, for which we coin ``networklet" to represent a set of VMs that collaboratively provide some PaaS or SaaS service. Through extracting networklets from tenant requests and thus sharing them between multiple tenants, we can achieve a win-win situation for datacenter providers and tenants. Extensive evaluations show that, the proposed model and deployment algorithm indeed improve DCN resource utilization while maintaining performance guarantee.

In today's datacenters, resource requests from tenants are increasingly transforming into hybrid requests that may simultaneously demand IaaS, Paas, and SaaS resources. This paper tackles the challenge of modeling and deploying hybrid tenant requests in datacenters, for which we coin "networklet" to represent a set of VMs that collaboratively provide a PaaS or SaaS service. Through extracting networklets from tenant requests and thus sharing them between tenants, we can achieve a win-win situation for datacenter providers and tenants.

This paper proposes a new manifold learning method called "Soinnmanifold". Traditional manifold learning method needs a lot of computation and appropriate priori parameters. This has somewhat restricted the domains in which manifold learning can potentially be applied. However, with the high-dimensional inputs, our method can generate a lowdimensional manifold in the high-dimensional space and determine the intrinsic dimension automatically. Then we will use this manifold to do dimensionality reduction quickly. Experiments demonstrate that our method can get promising results with less time and memory.