Efficient collaboration among Unmanned Aerial Vehicles (UAVs) has significant performance improvement for UAV-based applications. Task offloading is the typical collaboration form for UAV system. However, it still be a challenging problem for UAV system due to task dependencies and the UAV mobility which makes the traditional offloading approaches inefficiency. In this paper, we model the offloading problem as the Sequential Task Offloading Problem (sTOP), which takes the task spatiotemporal dependencies into account. We propose a Graph-based Spatiotemporal Reinforcement Learning (GSTRL) framework, where the environment is modeled as a heterogeneous graph to capture the diverse relationships among system entities. A spatiotemporal state extraction module is designed, which integrates a Heterogeneous Graph Neural Network (HGNN) for spatial dependency modeling and a Long Short-Term Memory (LSTM) network for temporal dynamics. Based on the extracted representations, a masked Proximal Policy Optimization (mPPO) algorithm is proposed to make valid and efficient offloading decisions under multiple system constraints. Extensive experiments using real UAV trajectory and building distribution datasets validate that the proposed method improves the average reward by approximately 25% over state-of-the-art DRL-based and heuristic baselines, by increasing task success rate and operational effectiveness ratio (OER) to 30–50%, while reducing execution time by up to 40% in complex multi-UAV systems.

Low Earth Orbit (LEO) satellite constellations are crucial for enabling global coverage and low-latency services. Satellite Edge Computing enables on-board processing to improve task responsiveness and reduce backhaul load. Microservices, with their modular and lightweight design, naturally fit the dynamic and constrained satellite environment. However, deploying microservices on LEO satellites faces unique challenges, including stringent energy constraints, highly dynamic connectivity, and complex service dependencies across mobile nodes. In this paper, we propose a Microservice Deployment Framework for Satellite Edge Computing (SEC-MDF). We construct a microservice deployment model that includes real track dynamics, service dependencies, resource constraints, and data transfer patterns, and design a deployment strategy based on deep reinforcement learning (DRL). This strategy is augmented by Heuristic-based Episodic Reward Optimization (HERO), a tailored reward optimization mechanism for SEC-MDF. By integrating A*-based heuristic cost estimation, stage-aware episodic reward buffer, and adaptive reward normalization, HERO significantly enhances the performance of the deployment. Extensive experiments against both heuristic algorithms and DRL variants demonstrate that our framework reduces end-to-end latency by 40 % and system energy consumption by 29 %, significantly outperforming existing approaches. Ablation studies further confirm the critical contribution of the HERO mechanism to overall performance.

Cross machine traffic caused by distributed microservice deployment in edge computing significantly affects service performance. And the dynamics of the edge computing, such as fluctuating user request patterns and different network delay make static scheduling stategies challenging. To address these two issue, we first propose a Cross-Machine Traffic-Aware Scheduling Algorithm (CTSA), which models the microservice deployment process as a Markov Decision Process and utilizes a Dueling DQN-based approach to minimize cross-machine traffic while balancing node resource usage. Furthermore, we propose a Dynamic QoS-Aware Scheduling Framework (DQSF) that adapts deployment decisions in real time based on system monitoring to address the challenge of dynamics. Experimental evaluations using a real-world microservice application show that our approach significantly reduces request response time up to 30.2%, improves throughput to 36.7% and ensure Quality of Service (QoS) under dynamic edge computing continuum.

In multi-user edge intelligence system, achieving efficient and stable DNN inference under heterogeneous task priorities is a critical challenge. This paper presents a priority-aware edge-device collaborative inference scheme that models the entire inference workflow while explicitly incorporating user-level priority constraints. The optimization objective is twofold: to maximize, in priority order, the number of users with stable local queue under the strict constraint of server queue stability, and to minimize the total end-to-end latency across all users. A key component is the Server Queuing Latency Estimation (SQLE) algorithm, which decomposes the latency contributions of user priority interactions and iteratively estimates task queuing times. Compared with classical models such as M/D/1, SQLE achieves significantly higher accuracy under dynamic workloads. Based on the predicted latency, we further develop a two-stage offloading decision algorithm: Maximum User Prioritized Selection for Local Queue Stability (MUPS) determines the maximal subset of users whose local queues can be stabilized, and Fine-grained partition point for Latency Optimization (FOPL) refines offloading points to minimize global latency. Experiments on a heterogeneous edge-device system show that SQLE consistently achieves over 90% estimation accuracy with low error variance across varying system scales, significantly outperforming classical queuing models. Under different load, MUPS and FOPL supports more stable users and reduces average end-to-end latency, demonstrating its robustness and superiority over state-of-the-art methods.

Unmanned aerial vehicles (UAVs) have attracted widespread attention in post-disaster search and rescue (SAR) due to high flexibility and low-cost advantages. However, traditional centralized control approaches face problems such as poor adaptability and low robustness in complex and dynamic post-disaster environments. In order to improve the autonomous and execution efficiency of UAV cooperative search tasks, decentralized control methods have gradually become a research focus. However, how to efficiently realize autonomous path planning for UAVs under the condition of limited computational resources is still a key challenge to be solved. In this paper, we propose a dynamic adaptive path optimization method based on evolutionary neural network (DAPO-ENN), which combines the global search capability of evolutionary algorithms with the adaptive characteristics of neural networks to realize the centerless autonomous path planning and search coverage optimization of UAVs in post-disaster environments. DAPO-ENN can optimize the performance of the model under the limitation of computational resources, and adapt to the dynamic changes of the environment by online path optimization adjustment, so as to effectively improve the coverage efficiency while ensuring a high search coverage rate. The experimental results show that the DAPO-ENN proposed in this paper has stronger environmental adaptability and lower resource consumption than the existing comparison algorithms. The results suggest that the method provides an efficient and flexible solution for the cooperative search of UAVs after disasters.

We investigate the problem of path planning and task offloading for UAV clusters in a UAV-assisted edge computing scenario. UAVs autonomously make decisions regarding path planning, continuous service provision, and task offloading based on collected information. In this setting, terminal equipment (TE) cannot directly connect to servers; thus, UAVs act as both edge servers and communication relays, proactively providing services to TEs. We construct a fine-grained temporal scale model that decomposes UAV actions into atomic time units, transforming decision-making on specific behaviors into state transition decisions. This approach better accommodates the needs of time-varying environments. Regarding path planning, given the time-sensitive nature of TE requests, we focus on how to provide stable and timely computational services to TEs. We propose a multi-agent reinforcement learning algorithm capable of dynamically sensing task priorities to enhance Quality of Service (QoS), with optimizations made in terms of task completion rate, UAV energy consumption, and processing delay. About task offloading decision, we introduce a dual-keyword-based offloading algorithm to optimize the binary offloading process. Finally, we conduct simulation experiments to demonstrate the effectiveness of the proposed algorithms, and comparative experiments confirm their superiority.

To enable edge AI providers to recruit edge devices and use them to deploy AI models and provision inference services, we conduct a comprehensive mathematical and algorithmic study on a novel incentive and optimization mechanism based on repeated auctions. We first model and formulate a time-cumulative social cost optimization problem to capture the challenges of the trade-off between cost and accuracy, the dependency between adjacent auctions, and the need of achieving desired economic properties. Then, to solve this intractable non-linear integer program in an online manner, we design a set of polynomial-time algorithms that work together. Our approach dynamically chooses and switches winning bids under careful control, incorporates online learning to overcome posterior inference accuracy and workload queue dynamics, and leverages randomization to strategically convert fractional decisions of model placement and query dispatch into integers. We also allocate payments to meet the necessary and sufficient conditions for the desired economic properties. Further, we rigorously prove the constant competitive ratio, the sub-linear regret and fit, and the truthfulness and individual rationality for our proposed approach. Finally, through extensive experiments using real devices, AI models, and data traces, we have validated the substantial advantages of our proposed approach compared to the baselines and the state-of-the-art methods.

In edge computing, end devices (EDs) containerize tasks with the necessary resources and offload subsets to a nearby high-capacity edge server (ES) to improve efficiency. Most existing research focuses on inseparable task offloading to minimize response times or resource allocation to reduce energy consumption. However, task execution can be speeded up with excessive computing and network resources, it will increase energy consumption and incur unnecessarily high costs. Besides, complex applications like autonomous driving often partition sequential tasks to improve performance, necessitating a joint optimization of sequential task offloading and multi-resource allocation. In this paper, we introduce a Stackelberg game-based framework to model the interplay between these elements.EDs, acting as leaders, determine the offloading breakpoints of sequential tasks and the locality for processing. TheES, as the follower, uses the Karush-Kuhn-Tucker (KKT) conditions and a Boundary-constrained quasi-Particle Swarm Optimization (Bc-qPSO) algorithm to refine computing and network resource allocation, aiming to reduce system costs effectively. Our simulations show that the proposed algorithms reduce cost by approximately 10%-20% compared to traditional methods, highlighting their potential for improving the efficiency of edge computing systems.

In dynamic edge environments, many nodes are constantly changing their geographical locations, leading to unstable network links and even disconnections among nodes relying on wireless communication. Existing research mainly considers optimizing resource allocation and task processing delays, ignoring the impact of transmission delays in fluctuating network environments. However, the transmission time of tasks or requests may be longer than the processing time, which can significantly affect the quality of service for delay-sensitive services. In such scenarios, the static strategies of EdgeMesh always perform terribly under unstable network. This makes the transmission time of task data fluctuate frequently, resulting in an increase in response time. Therefore, we proposed an adaptive strategy based on network-aware to improve the transmission delay. We implemented such a strategy by developing a proxy named NAP to assist EdgeMesh in dynamic traffic management. In order to adapt to the complex and volatile conditions of edge networks, NAP monitors every node in real-time. It determines the feasible path with the shortest expected transmission time based on the size of different task data, network and resources status among nodes, then proxies the traffic and forwards it in the kernel state. Numerous experimental results demonstrate that NAP can reduce task transmission time by $50 \%$ and significantly improve the end-to-end request latency in edge computing environments compared to native strategies of EdgeMesh.

Deep Neural Network (DNN) has been widely used in intelligence applications due to its excellent performance in executing inference tasks. Since DNN tasks require a large amount of computation and high-resolution raw input, collaborative inference is proposed to partition the DNN model from the middle layer to minimize the end-to-end latency. However, when existing work attempts to combine collaborative inference with task offloading in cloud-edge-end systems, two challenges result in poor latency and throughput: excessive layers in the model make it difficult to find suitable partitions, and multiple decision variables make the reinforcement learning agent challenging to converge. This paper aims to reduce the long-term average end-to-end latency of DNN tasks by jointly optimizing task offloading, model partitioning, and resource allocation in dynamic environments. To solve the problem of failing to find suitable partitions, we propose a novel Optional Partition Point Compression (OPPC) algorithm, which selects the high-quality partition points based on layers’ output feature to reduce the difficulty of model partitioning. To improve the convergence performance of the agent, we propose a Reinforcement Learning based Collaborative Inference Optimization (RLCIO) algorithm. Unlike the existing reinforcement learning architecture, RLCIO decouples resource allocation using an Edge Computing Resource Allocation (ECRA) algorithm to reduce the agent’s decision variables. Simulation results show that the RLCIO algorithm performs better than five related schemes, reduces the average end-to-end latency of system tasks by 72% and improves the system throughput by 3.5x in the best case.

KubeEdge is a widely used edge computing system. It ensures quality of service through multi-replica deployment, horizontal scaling, etc. EdgeMesh, provided by KubeEdge, can offer key functions such as edge service discovery, cross-subnet communication, load balancing, etc. However, the native load balancing policies provided by EdgeMesh do not consider the wide geographic distribution of edge nodes. It forwards user requests to each application replica on all nodes, leading to higher end-to-end latency and lower throughput.Therefore, we propose a multi-level forwarding (MLF) policy based on EdgeMesh. The MLF divides the edge nodes close to each other into the same cluster. Request forwarding within the same cluster is called intra-cluster forwarding while forwarding between different clusters is called extra-cluster forwarding. It prioritizes processing user requests locally, then considers forwarding the requests to another node to process in the intra-cluster, and finally selects the best extra-cluster node for forwarding and processing. Our experimental evaluation results show that MLF can effectively reduce the end-to-end latency and improve the application throughput in edge computing scenarios compared to EdgeMesh’s default load balancing policies.

Intelligent terminal devices (TDs) usually request delay-sensitive and resource-demanding jobs, which are consisted of many sequential tasks. Mobile edge computing (MEC) offloads tasks to edge networks closer to TDs, making up for the lack of long delay response in the cloud, but it has a limited energy supply. Thanks to low-energy TDs also having processing capacity, it is a critical and challenging issue to offload sequential tasks for sustainable computing and reducing carbon emission in aterminal-edge-cloud(TEC) architecture. Existing research on offloading is limited to MEC orcloud-edgecoordination environment, and ignores the impact of sequential task (S-Task) constraint and service constraint. To bridge the gap, our paper first formulates the jointly optimalS-Taskoffloading and service deployment (JOTOSD) problems objected to maximize the energy utility related to response delay, which is NP-hard and is divided into deployment and offloading sub-problems. Then, we propose a comprehensive offloading and deployment (COD) method, including the Break-Point (BP) algorithm and the convex programming-based edge offloading (CVEO) algorithm under a service deployment strategy provided by an iterative service deployment (ISD) algorithm. Simulate results prove that the proposed method can improve by about 20% of energy utility by compared with other heuristic algorithms.

Loosely coupled and highly cohesived microservices running in containers are becoming the new paradigm for application development. Compared with monolithic applications, applications built on microservices architecture can be deployed and scaled independently, which promises to simplify software development and operation. However, the dramatic increase in the scale of microservices and east-west network traffic in the data center have made the cluster management more complex. Not only does the scale of microservices cause a great deal of pressure on cluster management, but also cascading QoS violations present a substantial risk for SLOs (Service Level Objectives). In this paper, we propose a Microservice-Oriented Topology-Aware Scheduling Framework (MOTAS), which effectively utilizes the topologies of microservices and clusters to optimize the network overhead of microservice applications through a heuristic graph mapping algorithm. The proposed framework can also guarantee the cluster resource utilization. To deal with the dynamic environment of microservice, we propose a mechanism based on distributed trace analysis to detect and handle QoS violations in microservice applications. Through real-world experiments, the framework has been proved to be effective in ensuring cluster resource utilization, reducing application end-to-end latency, improving throughput, and handling QoS violations.

Service Mesh has become one of the most popular microservices governance frameworks. In the service governance function of traffic management, how to design an efficient strategy for request distribution to minimize response time has attracted wide research interest. Generally, a traffic policy can be pre-set manually. However, such a static policy always performs terribly when the load changes dynamically. That is, Service Mesh cannot adaptively select the optimal policy when setting service version diversion and selecting instances on the service subset. To address this issue, in this paper, we designed an adaptive strategy based on split-flow to achieve global load balancing. Furthermore, we implement such a strategy by developing a plug-in named DTMA for Istio, a typical representative of Service Mesh. By obtaining the Service Mesh node’s topology and global instance load distribution in real-time, Istio updates the optimal split-flow weight and selects the suitable load balancing strategy. Extensive experimental results show that our algorithm can reduce the response time by nearly half and keep a load of nodes and instances stable compared with Istio’s native configuration. The optimization effect of minimizing the response time is achieved through real-time load balancing.

Edge computing provides an opportunity to improve the quality of service (QoS) of Artificial Intelligence (AI) apps for the Internet of Things (IoTs) scenarios. It is an important way to improve the QoS of intelligent apps by deploying Deep Neural Network (DNN) models on edge nodes. Though the DNN execution time affects the QoS of apps significantly. Due to the limited and dynamic edge resources, and sudden load to edge nodes, it is hard to guarantee the DNN execution efficiency. In this paper, we conduct fine-grained decomposition of DNN tasks and propose a Cloud Edge Collaborative Dynamic Task Scheduling mechanism based on DNN layer-partitioning technique. The approach can realize the collaborative computing of DNN models between cloud and edge, and improve the execution efficiency of DNN models, which guarantees the QoS of AI apps. Through simulation experiments, compared with the existing task scheduling mechanism and AI app deployment mode, we show that the proposed cloud edge collaborative dynamic task scheduling mechanism can effectively reduce the average service response time in the edge intelligent system, so as to improve the apps' overall QoS of the system. Meanwhile, the task scheduling mechanism designed in this paper makes it possible for more complex intelligent models to run in a resource-constrained edge environment.

The Non-Volatile Memory (NVM) is the promising device to store data and accelerate big data analysis due to its excellent I/O performance. However, we find that simply replacing Hard Disk Drive (HDD) with NVM cannot bring the expected performance improvement. In this paper, we take the data recovery issue in Hadoop File System (HDFS) as a case study to investigate how to take advantage of the performance of NVM. We analyze the data recovery mechanism in HDFS and find that the configuration of replication tasks in the DataNode can affect the data recovery significantly. We conduct extensive analysis and experiments to tuning the configuration and also get some interesting findings. With the new configuration, we increase the data recovery performance improvement from 17% to 71%. At the same time, we can also improve the execution performance of MapReduce tasks to 28% to 59% through optimized configuration.

The execution and completion of analytics jobs can be significantly inflated by the slowest tasks contained. Despite task replication is well-adopted to reduce such straggler latency, existing replication strategies are unsuitable for geo-distributed analytics environments that are highly dynamic, uncertain, and heterogeneous. In this paper, we firstly model the task replication and scheduling problem over time, capturing the geo-analytics features. Afterwards, we design an online algorithm, GeoClone, to select tasks to replicate and select sites to execute the task replicas in an irrevocably online manner, through jointly considering the execution progress of each job and the resource performance in each site. We rigorously prove the competitive ratio to exhibit the theoretical performance guarantee of GeoClone, compared against the offline optimal algorithm which knows all the inputs at once beforehand. Finally, we implement GeoClone with Spark and Yarn for experiments and also conduct extensive large-scale simulations, which confirms GeoClone's practical superiority over multiple state-of-the-art replication strategies.

The emerging Non-Volatile Memory (NVM) technology has given rise to an opportunity to accelerate big data analysis. In this paper, we investigate the joint job and data scheduling problem in private cloud data center with a hybrid storage system, and we propose Labeling Scheduler, a flexible labeling-based approach for jointly scheduling. The core idea of the approach is to introduce the labeling system to characterize the features of big data analysis jobs and data objects, and conduct data replacement dynamically between NVM and disk. To the best of our knowledge, this is the first work to introduce the labeling methodology to the big data analysis problem in the cloud data center with a hybrid storage system. We conduct extensive simulations and the simulation results show that the Labeling Scheduler has a significant improvement on system utility compared to the method without labeling information. In addition, the Labeling Scheduler guarantees a high NVM hit rate, which is valuable for NVM endurance enhancement.

In this paper, we investigate the joint job and data assignment problem in cloud data centers with non-volatile memory (NVM) for makespan minimization. Through extensive analysis, we find that there is an indicator variable that characterizes the hardness of the problem. Depending on the value of the indicator variable, we classify our problem into three cases: inf-case, opt-case, and nph-case. We first show that there is no feasible assignment under the inf-case. For the opt-case, we present an optimal algorithm. We show that a mixed data assignment with diversified popularity achieves high memory utilization. For the nph-case, we first prove the problem's NP-hardness and then propose a heuristic algorithm and a 2-approximation algorithm to tackle it. We conduct extensive simulations, and we find that the performance of the heuristic algorithm is better than the 2-approximation algorithm and that it is nearly the same as the theoretical optimal solution.

Flow scheduling is one of the primary issues in data centers. The efficiency of flow scheduling affects the user experience significantly, since the service latency is determined by the flow completion time (FCT). Current transport protocols are based on the Processor Sharing (PS) policy by dividing the link bandwidth equally. As short flows may be blocked by long flows under PS policy, these protocols could not meet latency requirements. The Shortest Remaining Processing Time (SRPT) policy provides a near-optimal solution in term of reducing average FCT. However, this flow scheduling policy may cause long flows suffering unfair delays. In this paper, we propose MERP (Minimizing Expansion Ratio Protocol) for flow scheduling, which aims to navigate the tradeoff between fairness and the average FCT. We propose expansion ratio as a new metric to describe the gap between reality and ideal in terms of acquired link resources by a flow. Larger expansion ratio indicates that the flow obtains fewer bottleneck link resources. We strive to make the flow scheduling relatively fair by reducing the largest expansion ratio of all the flows. And we further demonstrate that reducing expansion ratio is more reasonable than reducing average FCT for services with the Partition/Aggregate pattern. And to meet the scalability requirements, we implement distributed MERP through flow preemption and explicit rate control. The experiments indicate that MERP could significantly reduce the completion time for 99.9th percentile flows under high load.

As tenants take networked virtual machines (VMs) as their requirements, effective placement of VMs is needed to reduce the network cost in cloud data centers. The cost is one of the major concerns for the cloud providers. In addition to the cost caused by network traffics (N-cost), the cost caused by the utilization of physical machines (PM-cost) is also non-negligible. In this paper, we focus on the optimized placement of VMs to minimize the cost, the combination of N-cost and PM-cost. We define N-cost by various functions, according to different communication models. We formulate the placement problem, and prove it to be NP-hard. We investigate the problem from two aspects. Firstly, we put a special emphasis on minimizing the N-cost with fixed PM-cost. For the case that tenants request the same amount of VMs, we present optimal algorithms under various definitions of N-cost. For the case that tenants require different numbers of VMs, we propose an approximation algorithm. Also, a greedy algorithm is implemented as the baseline to evaluate the performance. Secondly, we study the general case of the VM placement problem, in which both N-cost and PM-cost are taken into account. We present an effective binary-search-based algorithm to determine how many PMs should be used, which makes a tradeoff between PM-cost and N-cost. For all of the algorithms, we conduct theoretical analysis and extensive simulations to evaluate their performance and efficiency.

Internetware challenges distributed systems in aspects from operating platforms, programming models, to engineering approaches, etc. Cloud computing based on virtualization is now a popular paradigm which can meet the dynamic resource allocation requirements of Internetware. Software entities dispersed on distributed nodes over the Internet, now are evolving into self-contained, autonomous software services. These software entities which are often deployed on virtual machines (VMs), are coordinated dynamically to achieve flexible design objectives. To improve the utilization of infrastructure resource, VMs processing components of a software should be consolidated to fewer physical machines (PMs). However, as the increasing trends of communication-intensive softwares, data traffic among VMs should be considered as well. And for the sake of safety and QoS (Quality of Service), certain VMs (e.g. backup nodes) are mutually-exclusive which means some VMs require to be placed on different PMs. In this paper, we investigate the online software placement problem with the target to minimize the network traffic cost, while taking into account the mutually-exclusiveness of VMs. We provide a formal problem description and its NP-hardness analysis. The proposed algorithm places the VMs that have heavy traffic on the same PM, while isolating the mutually-exclusive VMs simultaneously, which can guarantee high effectiveness and reliability for softwares in Internetware, respectively. The simulations show our algorithm reduces the traffic cost by 29% compared against the existing approaches.

Wireless sensor networks (WSNs) constitute an important area of research that is emerging. This is taking place due to the rapid and significant developments, which have led to sensing devices with increasingly smaller size, faster processing, lower energy consumption, as well as larger storage and communication capacities. In addition, as the amount of physical, chemical and biological conditions that are able to be sensed increases, WSNs are finding numerous applications in areas such as environmental, military, health care, and infrastructure monitoring. Many of these applications involve lining up the sensors in a linear form, making a special class of these networks, which are defined as Linear Sensor Networks (LSNs). In a previous paper, we introduced LSNs and provided a classification and motivation for designing networking protocols that can take advantage of the predictable linearity of the topology in order to optimize the performance, reliability, fault tolerance, energy consumption, and network lifetime. In this paper, we provide a topology discovery protocol for thick LSNs where, due to the nature of the monitored structure or area, and the deployment strategy, the nodes are assumed to exist between two parallel lines that extend for a relatively long distance compared to their transmitting range. As a result of the discovery process, a small percentage of the deployed nodes are selected to be a part of a backbone, which can be used for efficient communication between the other nodes in the LSN. The protocol takes advantage of the linearity of the network in order to reduce the amount of exchanged control messages, reduce energy consumption, and increase scalability. Two different strategies for topology discovery are presented, and simulated in order to verify and compare their operation, and efficiency.

As more and more services are offered in clouds, it is possible to meet the diverse demands of users via service composition. Selecting the optimal set of services, in terms of QoS, is the crucial issue when many functionally equivalent services are available. In this paper, we investigate the service selection problem under the service replica limitation constraint. The objective is to select the optimal service set which brings out the minimal response time. We estimate the communication latency with the network coordinate system. Based on the estimated latencies and service composition paradigm, our selection algorithms find the services which will result in low latency under various replica limitation constraints. We evaluate our approaches via extensive simulations, the experimental results of which show that our algorithms work efficiently.