Scheduling across cloud and edge systems must handle job heterogeneity, dynamic arrivals, and coupled resource constraints. To address the above issues, we present GT-MARL, a reinforcement learning framework with graph and Transformer enhancements for joint task selection and resource allocation across multiple clusters. At each decision epoch, GT-MARL encodes a heterogeneous graph over tasks and resources with Hierarchical Attention Network (HAN) to capture structural dependencies, applies a temporal Transformer to model backlog evolution and delayed interactions, and produces factorized decisions for each cluster through a Transformer-guided task selection head and discrete resource allocation with a Multi-Layer Perceptron (MLP) head. Training follows centralized training with decentralized execution (CTDE) using a centralized critic and masked action spaces. Experiments on both synthetic and trace-replay workloads show that GT-MARL consistently achieves the best p95 job completion time (JCT), while keeping mean JCT competitive with representative heuristic and learning baselines. Specifically, GT-MARL obtains 35.56 mean JCT and 70.76 p95 JCT on the synthetic workload, and 144.21 mean JCT and 264.47 p95 JCT on the trace-replay workload. On the synthetic workload, this corresponds to a 26.3% p95 JCT reduction relative to the best mean baseline with only a 3.0% increase in mean JCT. Additional queueing and service decomposition across workload intensities indicates that the tail-latency gain mainly comes from alleviating queue accumulation rather than shortening the inherent service time of tasks. GT-MARL provides a favorable tradeoff between mean latency and tail latency for scheduling under SLO constraints across cloud and edge systems.

Deploying Large Language Models (LLMs) on edge–cloud collaborative clusters leverages cloud elasticity and edge low-latency to mitigate resource bottlenecks. Such deployment, however, necessitates coordinated task offloading at the edge and resource scaling in the cloud. Deep Reinforcement Learning (DRL) methods like Proximal Policy Optimization (PPO) are unable to simultaneously learn strategies for both task offloading and resource scaling. While recent studies have introduced Multi-Agent Reinforcement Learning (MARL) methods for joint optimization, they are limited by the need for extensive expertise in specific cluster configurations to formalize the reward function, making them non-scalable. This paper proposes FGPPO, a Graph Neural Network (GNN)-based distributed multi-agent task placement strategy for the joint optimization of task offloading and cloud resource scaling. FGPPO comprises distributed agents responsible for task offloading and resource scaling, respectively, and employs a GNN-based reward model for distributed reward assignment. The proposed reward model leverages a heterogeneous graph-attention mechanism to holistically integrate service quality, cost, and interdependencies among multiple resource nodes. Experimental results show that FGPPO outperforms the baselines across various real-world traces and cluster configurations. The average cost of FGPPO is 1.25 times that of the offline greedy method, while it significantly reduces the SLA violation rate. These findings indicate that the proposed method holds promise for optimizing costs for LLM providers while ensuring user-side service quality.

As the complexity of business and scenarios contin-ues to grow, the traditional, inefficient, and cumbersome domain modeling process can no longer adapt to the rapid iteration requirements. In recent years, technological breakthroughs in generative artificial intelligence (Generative AI), particularly in large language models (LLMs), present novel opportunities to enhance domain modeling efficiency. While LLMs demonstrate baseline capabilities in information extraction, their potential for constructing complex domain-specific models remains underexplored. This study investigates how LLMs can facilitate automated domain modeling tasks and support domain modeling education. Through systematic experimentation, we evaluate the impacts of LLM fine-tuning techniques and prompt engineering strategies on model outputs, while comparing two distinct generation modes: indirect model construction via domain element extraction versus direct domain model generation. Our empirical results demonstrate that LLMs exhibit significant potential in supporting high-efficiency domain modeling, with fine-tuning techniques and indirect generation modes yielding superior out-comes. Furthermore, we illustrate LLMs' utility as pedagogical tools for domain modeling education while identifying critical limitations and implementation risks that warrant consideration in both practical applications and future research.

With the rapid development of cities and the increase in the number of motor vehicles, traditional intelligent traffic video analytics systems face a significant challenge due to soaring computational demands and limited network transmission resources. Today’s widely used cloud-based traffic video analytics system, which is generally reliant on transmitting all video data to centralized cloud servers, suffer from high latency during network fluctuations and inability to respond promptly to urban traffic management. This paper introduces a cloud-edge collaborative framework for traffic video analytics. Within this framework, edge servers serve as intermediaries between video sources and the cloud center. The framework prioritizes the offloading of computing tasks to nodes located near the video sources, rationally leveraging the limited computing and network resources to mitigate network transmission load. We develop a measurement-based analytical model to describe the trade-offs among network latency, inference latency, and analytics accuracy in edge-based real-time video analytics system. As a core element of our approach, we present a resolution selection and bandwidth allocation algorithm based on Lyapunov optimization and heuristic search, designed to dynamically adjust video resolution and distribute bandwidth between edge and cloud servers to balance latency and accuracy without requiring future information. Experiments on a cloud-edge collaborative video analytics system demonstrate the algorithm’s effectiveness in substantially enhancing accuracy and responsiveness.

With the continuous development in the field of machine learning, there is an increasing demand for the cloud-based machine learning inference services, which are latency-sensitive tasks, such as the service requests from the Internet of Things (IoT) devices. These inference services are generally accompanied by fluctuations and uncertainties, so they often require vastly varied numbers of servers at different time spots. As a result, how to dynamically and rationally schedule cloud servers for inference services has become an important issue. Alibaba Cloud currently provides its serverless instances called elastic container instance (ECI), and due to the advantages of pay-as-you-go billing and second-level elasticity, they are well-suited for handling bursty or fluctuating workloads. At the same time, Alibaba Cloud’s subscription-based elastic compute service (ECS) instances can be used for steady-state workloads. Our objective is to dynamically combine these two types of instances to deal with inference service requests. In this article, we propose a deterministic online algorithm that can rationally schedule these two types of instances to optimize costs without requiring knowledge of future workloads. We prove that the proposed online algorithm achieves a competitive ratio of no more than 2 compared to the optimal offline algorithm. Through simulation experiments, we demonstrate that our algorithm outperforms three benchmarks, which are all-reserved algorithm, all-on-demand algorithm, and traditional online algorithms that only use ECS instances. Our algorithm exhibits superiority across various workloads and can significantly reduces costs in most cases.

Fueled by the advancements in 5G networks, Internet of Things (IoT) technologies are undergoing rapid development. Utilizing the computational power and elasticity of cloud computing, deploying IoT applications as Software-as-a-Service (SaaS) can reduce costs while enhancing scalability. However, in recent years, large-scale cloud downtime caused by natural disasters and network attacks, which seriously affects SaaS applications, has been widely observed. Deploying SaaS applications across geo-dispersed instances in a distributed way can effectively avoid service termination caused by cloud downtime. Generally, SaaS applications face dynamic traffic, which can lead to variations in service qualities and potentially fail to meet user demands. Scaling cloud instances according to the real traffic could save costs for service providers while ensuring service qualities. Considering the lack of effectiveness of reactive-threshold-based scaling strategies, many works suggest proactive scaling strategies based on reinforcement learning (RL). However, most RL-based scaling strategies are cursed by dimensionality when facing dynamic traffic. Moreover, heterogeneous and time-varying multiregional instances often lead to a complex state space which intensifies the curse of dimensionality. To address the above issues, in this article, we propose an RL-based two-layer scaling strategy that addresses the challenge of cloud resource scaling in complex state spaces through hierarchical decision-making. We formulate the scaling of multiregional heterogeneous instances as a bi-objective optimization problem and discuss the Markov property of time-varying variables, including time-varying instances and dynamic traffic. Experimental results based on two real-world workload datasets show that our two-layer scaling strategy achieves more fine-grained, cost-effective scaling while maintaining service quality, compared to the baselines.

Service level agreements (SLAs) formulated by cloud rendering service providers and users are varied, as users may have diverse performance requirements for their own jobs. This leads to a complex issue that cloud resources need to be allocated to rendering jobs in an appropriate and effective manner to satisfy users' diverse SLAs. To address this issue, in this paper, we propose a novel fair and efficient resource allocation algorithm, which aims to maximize the execution efficiency of rendering service applications while satisfying users' diverse SLAs. Firstly, to satisfy users' diverse SLAs, we propose a rigorous definition ofweighted acceleration ratio fairness, whose guiding principle is that the execution speed of a rendering job should be proportional to its weight determined by users' SLAs. Then, under the guidance of the proposed principle of acceleration ratio fairness, we formulate a new algorithm to fairly allocate resources to rendering jobs. Lastly, to improve execution efficiency and coordinate efficiency and fairness, we propose a fair and efficient resource allocation algorithm with relaxing fairness in resource competitive and non-competitive situations separately for rendering service applications. With extensive experiments that involve real rendering application workloads, we validate the effectiveness of our algorithms in improving execution efficiency and satisfying users' diverse SLAs.

For cloud storage, edge caching not only reduces the latency of delivering data to requesters due to closer delivery distances, but also potentially brings cost-savings to cloud users due to cheaper edge resources. However, cost-effective caching requires the exact knowledge of future requests, which is hard for cloud users to obtain in advance. It raises a risk of incurring more costs due to unpopular redundant replicas if data is blindly cached at the edge. To overcome this challenge, in this paper, we propose an online cost-effective caching algorithm based on reinforcement learning to dynamically make decisions of edge caching and caching lifespan online without any knowledge of the future. Further, we propose several mechanisms to enhance the cost performance of the proposed algorithm, to make up for possible mistakes made at the beginning of the learning and avoid cost-ineffective periods in caching lifespans. We then analyze the performance of caching decisions given by the proposed algorithm with enhancements. Finally, we conduct extensive simulations driven by real-world traces under prevalent pricing schemes of both the cloud and the edge, which reveals that significant cost-savings and superiority over benchmark algorithms can be achieved.

Currently, with the popularity of live video applications, VSPs (video service providers) begin to use cloud servers to enhance user experience and reduce operational costs. In this paper, we consider VSPs leveraging a cloud-edge cooperative model to deliver video services for cost reduction. Since bandwidth costs make up a significant portion of VSPs' operating expenses, we mainly consider bandwidth cost optimizations in traffic allocation. In addition, the QoE (quality of experience) is also very important, while the latency has a larger impact on QoE. Thus our traffic allocation approach aims to strike a fine balance between minimizing bandwidth cost and bounding the latency experienced by clients. Such a trade-off is difficult to optimize with some prevailing bandwidth billing methods such as the$95^{th}$percentile bandwidth billing. We quantify such a trade-off by constructing a linear bandwidth cost optimization problem. We first describe the offline version of the optimization problem, and then design an online greedy algorithm that considers minimizing the current bandwidth cost at each time slot. By applying the Lyapunov optimization framework, we design another online algorithm based on the original greedy one. We prove that the time average delay achieved by our online algorithm is smaller than the upper bound we set when certain conditions are satisfied. Through extensive simulation experiments, we show that the proposed online algorithm can significantly reduce both the bandwidth cost and the time average delay of clients.

Hadoop, as an open-source implementation of the MapReduce paradigm, is increasingly being used in both industry and academia for large-scale data processing. Yarn, one of the core components of the second-generation Hadoop, manages cluster resources and job scheduling. Minimizing the total completion time of a set of MapReduce jobs is a point worth exploring in terms of Yarn’s performance. Hadoop’s default schedulers, including first-in-first-out (FIFO), Fair, and Capacity, do not consider the characteristics and preferences of job resource demand, resulting in insufficient resource utilization. Therefore, in this paper, a new job scheduler named Q-scheduler is proposed. It uses reinforcement learning (RL) to accumulate scheduling experience autonomously based on the Fair scheduler. Specifically, the proposed scheduler consists of a Classifier and a Decider. The Classifier classifies jobs through similarity measurement, and the Decider, as an agent with a Q-Table, considers the execution order of different job classes and updates the state-action values of the Q-Table to learn optimal scheduling. The experimental results show that Q-scheduler can reduce the total completion time of the job set and improve resource utilization.

Due to the advantages of ease of management, high elasticity, and low prices, serverless computing has gained more and more momentum in recent years. Serverless computing allows developing and deploying applications in the form of Function-as-a-Service (FaaS). Application service providers can benefit greatly from deploying and running their applications in public serverless FaaS cloud platforms. Currently, public FaaS cloud platforms generally offer multiple computation resource configurations for application service providers to select for deploying and executing their functions with a pay-per-use billing. Application service providers can select a function instance with a higher processing capability to achieve a shorter job completion time, but it also incurs a higher cost, while service providers often have budget constraints for running their jobs. Thus, when scheduling serverless application jobs in public FaaS cloud platforms, application service providers need to trade off between the job completion time and the cost, which is generally an NP-hard problem. To address these issues, in this paper we propose an improved genetic optimization algorithm for scheduling serverless application jobs in public FaaS clouds, which can achieve the minimization of the job completion time within the given budgets. Specially, during the evolution process of our genetic scheduling algorithm, we use a greedy strategy to search for optimal schedules. Through experiments running on both synthetic and real-world data, we validate the efficiency and effectiveness of our proposed genetic optimization based scheduling algorithm in minimizing the job completion time within budget constraints.

In the era of Industry 4.0, the widespread application of Internet of Things (IoT) technology enables us to monitor the operational status of production equipment through sensors and creates new requirements for equipment anomaly prediction and analysis. It can shift maintenance tasks from passive to proactive, reducing downtime and repair costs associated with equipment failures, and ensuring production safety and efficiency. However, the main challenge in industrial IoT applications lies in acquiring a sufficient amount of anomaly data. To address this issue, we propose a method that analyzes normal operating data of equipment to reveal trends in equipment status, combining multiple-model fusion prediction and multi-index coordinated decision-making. Firstly, a multi-model fusion strategy is adopted for prediction, integrating models such as XGBoost, LightGBM, and LSTM to enhance the accuracy of equipment attribute prediction. Secondly, through a multi-index coordinated approach, considering combinations of multiple indicators, an adaptive dynamic threshold rule is formulated based on the residuals between predicted and actual values to achieve early warning of equipment anomalies. Finally, we validate the effectiveness of this method using a mine ventilation fan as an example. The experiment demonstrated that this method can detect equipment anomalies one day earlier than traditional threshold alarm methods, achieving proactive equipment maintenance.

Currently, there are a number of users pursuing to purchase professional services for executing their jobs. Meanwhile, a service provider generally purchases on-demand or reserved instances from public IaaS (Infrastructure-as-a-Service) cloud platforms to elastically run users’ submitted jobs and then charges users for job-execution services accordingly. In the process of service provisioning, to maximize social welfare, service providers should systematically and economically decide their instance purchase, job scheduling, and service pricing schemes. For making optimal decisions, several challenges need to be addressed, including the online arrival of users, the NP-hardness of the problem, and the possible misreporting of selfish users. Faced with the above challenges, we design an online auction mechanism which can help service providers provisioning better services to achieve optimal social welfare, without requiring future information. Through strict theoretical analysis, we prove that the designed mechanism can guarantee truthfulness and individual rationality, run in polynomial time, and satisfy budget balance, while achieving competitive social welfare. The extensive evaluations on the basis of both synthetic and realistic Google cluster datasets demonstrate the effectiveness and efficiency of the designed mechanism.

Uncertainty exists widely in various fields, especially in industrial manufacturing. From traditional manufacturing to intelligent manufacturing, uncertainty always exists in the manufacturing process. With the integration of rapidly developing intelligent technology, the complexity of manufacturing scenarios is increasing, and the postdecision method cannot fully meet the needs of the high reliability of the process. It is necessary to research the pre‐elimination of uncertainty to ensure the reliability of process execution. Here, we analyze the sources and characteristics of uncertainty in manufacturing scenarios and propose a meta‐model architecture and uncertainty quantification (UQ) framework for uncertainty modeling. On the one hand, our approach involves the creation of a meta‐model structure that incorporates various strategies for uncertainty elimination (UE). On the other hand, we develop a comprehensive UQ framework that utilizes quantified metrics and outcomes to bolster the UE process. Finally, a deterministic model is constructed to guide and drive the process execution, which can achieve the purpose of controlling the uncertainty in advance and ensuring the reliability of the process. In addition, two typical manufacturing process scenarios are modeled, and quantitative experiments are conducted on a simulated production line and open‐source data sets, respectively, to illustrate the idea and feasibility of the proposed approach. The proposed UE approach, which innovatively combines the domain modeling from the software engineering field and the probability‐based UQ method, can be used as a general tool to guide the reliable execution of the process.

Nowadays, the Internet is rapidly evolving toward the future of the Internet of Things (IoT), where billions or even trillions of edge devices may be interconnected. The proliferation of network cameras and the advancement of IoT technologies have provided broader opportunities for data collection and utilization. In the past, the massive real-time videos generated by network cameras were mostly transmitted over the network to the cloud for analysis. However, due to network speed limitations, the latency incurred by uploading all videos to the cloud makes it difficult to meet the real-time requirements of video analysis. While edge computing significantly reduces latency, the computational capabilities of edge devices are limited, making it difficult to handle large amounts of real-time video data. In this article, we introduce a real-time video processing framework called DeepVA, which utilizes cloud-edge collaboration technology to reduce latency in real-time video processing and enhance the accuracy of analysis. The DeepVA framework incorporates the DRLVA video frame distribution algorithm based on deep reinforcement learning (DRL), which dynamically determines whether to distribute video frames for processing at the cloud or edge. To evaluate the performance of the proposed DRLVA algorithm, we first verify that it is superior to several other DRL-based distribution algorithms on the Gym environment. We also evaluate the performance of DeepVA on the MOT2015 data set, MOTSynth data set, and real campus surveillance videos. The experiments show that our DeepVA outperforms both cloud-only and edge-only solutions in terms of reducing latency and improving accuracy.

With the development of 5G network and edge computing technology, service providers can deploy applications on the edge cloud close to the users to improve service quality and efficiency. However, due to the limited number of edge resources and the dynamic change of user requests over time, it remains challenging for service providers to attain optimal resource procurement and job processing decisions at an edge data center. The computing capacity of public cloud resources is generally unlimited, but it is difficult to achieve low-latency performance similar to the edge resources and job transmission also requires high network bandwidth costs. In this article, we consider service providers using the edge-cloud collaborative mode to take advantage of the low-latency characteristics of edge resources and improve their scalability. Service providers purchase resources at the cloud and edge, respectively, to deploy applications, so as to satisfy dynamic service requests with different service quality requirements. In order to optimize the cost of service providers while ensuring the quality of their services, we propose a two-timescale online scheduling algorithm based on the Lyapunov optimization, which makes optimal decisions without knowing future information about users’ jobs. By determining the number of edge resources at a large time scale and the number of cloud resources at a small time scale, our algorithm can avoid frequent application deployments at edge while achieving rapid service response in a cost-effective manner. Rigorous theoretical analysis and extensive experiments based on both the synthetic and real-world data verify the effectiveness of our proposed algorithm.

Nowadays, many dynamic recommendations still suffer from the insufficiency of finding user online interest evolving patterns because of those complicated interactions. In general, each interaction is usually impacted by multiple underlying reasons, which needs us to open the “box” of each interaction instance instead of simply treating them as a pair-wise link. Besides, different users usually perform differently for their long-term and short-term tastes, leaving traditional sequential models far from personalized. In this article, we propose a novel recommendation model based on Graph Diffusion and Ebbinghaus Curve. Specifically, to explore the underline reasons for different interactions, we explore an underlying sub-graph for each interaction and find important reasoning paths within the sub-graph via a well-designed graph diffusion method. To capture users’ personalized strategies on long-term and short-term tastes, we are inspired by the Ebbinghaus Curve, which can naturally describe users’ memory patterns, and design an effective neural network to process users’ evolving behaviors. We conduct extensive experiments on four real-world datasets and the results further confirm the superiority of our model compared with existing state-of-the-art baselines.

In many fields, such as social networks and recommendation systems with high time requirements, fake news and false information are often released in real time, impacting on people’s daily life. Entity alignment (EA) in temporal knowledge graph (TKG) can fuse the information contained in entities by finding equivalent entities, thus helping to determine the regular pattern of disinformation under time change. The existing methods either ignore the use of temporal attributes’ information and structural information or the modeling of that is insufficient, which has become a major obstacle to the further and wider application of TKG EA. In this article, we put forward a new idea of training for the processing of time attributes and relational structure information, to further enhance the ability in the EA process of TKGs. By forming box embedding matrix and name embedding matrix, and adaptively fusing the above information, we propose a new TKG EA solution. We carry out comparative experiments on standard news media and social media datasets collected from the real world, which validates the effectiveness of our proposal.

Real-world heterogeneous information networks (HINs) are modeled as heterogeneous graphs, in which features and structures are often incomplete. Existing models employ manual imputation or dynamic adjustment to populate the incomplete data. However, there are some limitations in incomplete heterogeneous graph representation learning: 1) using populated data may lose content and high-level interaction information of HINs, even lead to negative impacts on the performance of downstream tasks; and 2) existing models fail to utilize the high-order heterogeneous structures in original incomplete network data. To resolve the above issues, in this article, we proposed a lattice-based incomplete heterogeneous structural attention network (LIHAN) for learning incomplete heterogeneous node embeddings. LIHAN first constructs characteristic lattice and structure lattice by mining characteristic sets and structure sets according to the partial order relations in between. Then, an improved lattice-based heterogeneous dual-attention mechanism is used to learn the heterogeneous node representations. Extensive node classification experiments are conducted on five open datasets to verify the superior performance of the proposed LIHAN model over the state-of-the-art models. Experimental results illustrate that LIHAN outperforms other methods on the micro-F1 and macro-F1 in node classification tasks. Moreover, experiments on different levels of lattices and the parameter sensitivity analysis shows the great stability during the process of experiments.

Edge storage is emerging as a novel storage paradigm, which offers the advantage of low latency and low cost, but has the disadvantage of limiting the scope of services to a certain area. In contrast, cloud storage offers anywhere services but has disadvantages in terms of latency and cost. In this paper, a collaborative storage scheme is proposed to leverage their complementary advantages. Considering the impact of collaboration on the cloud and the edge, we propose collaborative optimization models for cost and latency. To address the challenge of requiring future information when performing optimization for cost and latency, we first transform the long-term optimization into individual optimizations in each time slot using the Lyapunov optimization technique, based on which our algorithm decides whether a replica should be created at the edge and which cloud tier a replica should be migrated to. Then, we prove that the proposed algorithm can obtain near-optimal costs and guaranteed latencies. Finally, we conduct extensive simulations driven by real-world traces, and show that our algorithm can achieve a trade-off between cost and latency and outperform other benchmark algorithms.

Root cause localization in microservice systems refers to finding the root cause that causes system anomalies using system information. Many methods construct a graph structure and perform random walk on it to localize the root cause. This is not suitable for larger systems due to the high computational overhead. Besides, the constructed graph is usually static which mismatches with evolving metrics. Different metrics also contribute differently to determining root cause. To address these challenges, we have developed DGERCL, a novel method that employs dynamic graph embedding to localize root causes in microservice systems. We construct a dynamic graph where nodes, edges, and features correspond to microservices, invocations, and metrics. DGERCL first gets invocation information by aggregating node embedding and features via a trainable structure. An LSTM then processes invocation information to update node embedding. We also propose a neighbor information aggregation method to enrich structure information and a self-attention-inspired mechanism to leverage the importance of metrics for better mining metrics information. Finally, a classifier maps node embedding learned by LSTM to possibilities belonging to root cause. We conduct comprehensive experiments on two microservice benchmarks. Our model achieves good results which demonstrates the effectiveness of DGERCL.

As increasingly complex functions are implemented in applications, directed acyclic graphs (DAGs) are widely used to model the inter-dependencies between individual functions. Cloud-based data processing platforms need to consider the complex topology of DAGs and arbitrary deadlines given by users for job scheduling, leading to an NP-hard decision-making problem. Leveraging spot instances in data processing platforms can achieve significant cost savings, but the unpredictable interruption of spot instances makes the problem of VM scaling and job scheduling more difficult. In this paper, a Reinforcement Learning (RL) based approach called SpotDAG is proposed to solve the auto-scaling problem for jobs modeled as DAGs on a data processing platform where spot instances are introduced. SpotDAG makes cluster scaling and job scheduling decisions at the same time by mapping its output to several meta-policies. This paper introduces the self-attention mechanism for feature extraction to help the intelligent agent learn faster. A mask layer after the output of the proposed RL-based algorithm circumvents illegal actions to ensure that a job is completed by its deadline. Extensive experimental results show that the proposed approach can significantly reduce the cost of instances for data processing platforms while ensuring that jobs are completed in time.

Communication networks represent communication between entities like social networks and microservice call graphs of microservice systems. Link prediction is useful in various communication network service systems such as predicting the relation between two services. Continuous-time dynamic graph (CTDG) is one form of representing temporal information in communication networks that treats them as a set of events occurring over time. Dynamic graph embedding for CTDG handles these events and disseminates event information to other nodes to get node embedding. Though dynamic graph embedding is suitable for making link prediction in communication networks, graph embedding for CTDG faces challenges such as how to model the event information dissemination process including information decaying over distance and influence of time information. To cope with this issue, we propose a CTDG-based dynamic graph embedding framework for dynamic communication networks link prediction called CTDGNN (Continuous-Time Dynamic Graph Neural Networks). In particular, we propose a self-adaptive information dissemination strategy based on node importance to update node embedding by disseminating event information. Finally, extensive numerical experiments on three real-world communication network datasets validate the effectiveness of our proposed model compared to other related methods.

The cloud storage boom has prompted providers to offer two storage tiers, i.e., hot and cold tiers, which are respectively purpose-built to provide the lowest cost for frequent and infrequent access patterns. However, for cloud users, it is non-trivial to determine cost-effective tiers because it is hard to obtain future access patterns in advance and is difficult to predict them exactly. The lack of future information poses a risk of increasing costs instead of saving costs. This is not the only challenge encountered when it comes to cost optimization. In this article, we take Amazon S3 as an example to analyze the pricing of two-tier cloud storage and derive several major challenges faced by cost optimization. Then, assuming a priori knowledge of future access patterns, we propose an optimal offline algorithm based on dynamic programming to determine cost-effective tiers for each time slot. Further, to handle online workload arrivals, we formulate the problem using Markov decision processes and propose RLTiering based on deep reinforcement learning. Eventually, the cost performance of RLTiering is evaluated based on real-world traces and prevalent Amazon S3 pricing, and the results show that it achieves significant cost-savings.

Nowadays, the improvement of the data acquisition capability of the Industrial Internet of Things (IIoT) systems has brought about higher data throughput. Users can deploy industrial big data analysis applications on cloud platforms in a pay-as-you-go way to deal with the unstable data generation and data analysis workloads in the IIoT scenario. To procure stable and flexible computing resources, as a traditional pay-as-you-go cloud instance procurement option, on-demand instances are widely used, but their expensive prices also significantly increase users’ cost burden. To reduce the data analysis cost, in this article, we establish a per-job cost-effective framework adopting spot instances, on-demand instances, and cloud storage for industrial big data analysis applications. In our framework, we take advantage of spot instances, which are computing instances provided at low prices under a pay-as-you-go model, to achieve cost savings. However, using spot instances carries the risk of being interrupted. Therefore, we propose to use a checkpointing mechanism to back up intermediate results to cloud storage to reduce the potential loss caused by spot instance interruptions. Considering the time sensitivity of industrial big data analysis applications, we use on-demand instances as alternative computing resources after spot instance interruptions to ensure that users’ jobs can be completed without high time latencies. Evaluation results show that our framework can achieve cost savings as well as minimize time latencies for users’ jobs.

The Amazon cloud platform sells its idle resources to cloud users as Spot Instances, which provide an ultra-low discount compared to the price of on-demand instances. Unlike the pricing strategy of the on-demand and reserved instances that use fixed price, the price of Spot Instances is dynamically changed, which introduce an interesting research topic of price prediction. In this paper, we firstly analyze the actual price distribution on a 90 days Amazon spot price history data downloaded from the Amazon Cloud platform, by using the parameter k-AMSE to represent the price fluctuation of a spot instance which can reflect recent data fluctuations better than MSE(mean square error). Then, We analyzed the factors that affect the price fluctuation and presented a prediction model based on the GRU(Gated Recurrent Unit) network. We compare the proposed algorithm with others and evaluate it with RMSE (root mean square error) measurement. The experiment results show that the GRU network approach can perform over others with an accuracy rate of 1.58e-3.

The recommendation system can recommend information to users efficaciously, which helps many users to obtain information in different fields. The paper recommendation is a research topic to provide authors with personalized papers of interest. However, most existing approaches equally treat title and abstract as the input to learn the representation of a paper, ignoring the author's interest and structure information of the academic network. In the paper recommendation system, authors and papers and the interaction of their information have a crucial impact on the efficiency and accuracy of the recommendations. However, most recommendation systems are usually designed based only on users. Therefore, we propose a method based on the author's periodic interest and academic graph network structure to obtain as much effective information as possible to recommend papers. Extensive offline experiments on large-scale real data show that our method outperforms the representative baselines.

For solving the problem of high bandwidth costs and service delays faced by content service providers (CSPs), edge computing services can be used as a supplement to existing cloud data centers for building more efficient Content Delivery Networks (CDNs). When there are a large number of requests for a content in one certain area, the content service provider can choose to rent an edge computing service near this area to lower the bandwidth cost for content delivery. But if the requests then drop after that, additional costs will be incurred instead due to the edge service renting. Therefore, it is necessary to dynamically decide whether to rent an edge service according to the request arrival situations in the future, but the future is often difficult to predict. For dealing with this problem, we propose an online edge service renting approach, as well as a corresponding request redirection algorithm, which can help content service providers save bandwidth cost significantly, while without requiring any knowledge about the future. Through theoretical analysis, we prove that the cost achieved by our online algorithm won't exceed 2 - α times compared to the optimal offline algorithm, where α is the bandwidth discount between edge and cloud services. Finally, by conducting extensive simulations with both real-world and synthetic data, we verify that our online edge service renting approach can effectively save costs for CSPs.

Storage-as-a-Service clouds generally offer both hot and cold storage tiers with different pricing options. Hot tiers provide a higher storage price but a lower access price, and vice versa for cold tiers. Many studies show that those user-generated data generally receive relatively high access frequency in the early period of their lifetimes while the overall trend of accesses is downward. Thus, when such kinds of data are hosted in clouds, they can be stored in hot tiers initially and then migrated to cold tiers for optimizing costs. However, the cost ofmigrationis non-negligible, and the number of accesses may then unexpectedly increase after migration, which indicates that a rash migration will incur more costs instead of cost-savings. For making optimal migration decisions, future data access curves are needed, but it is generally very hard to predict them precisely. To solve this problem, in this paper we propose a randomized online algorithm to optimize costs for those user-generated data stored in clouds, without requiring any future information. We show theoretically that the proposed algorithm can achieve a guaranteed competitive ratio of$1 + \frac {2(1-\lambda)}{e - 3 + 2\lambda + \lambda /\alpha }$, and it can be easily extended with prediction windows when short-term predictions are reliable. Eventually, we validate the effectiveness of our proposed algorithms through simulations driven by real-world video-visiting traces collected from a well-known video-sharing website.

An air traffic network is a special and complex Spatio-temporal network. What makes it unique is that multi-data sources-including airports, airlines and air routes-spatial dependence and strong temporal dependence in a dynamic environment. In this paper, we use big service to predict the flight departure delay time in air traffic networks. In the local services layer, we use graph sequences to model the Spatiotemporal network from multi-data sources, what is, using graphs to model the spatial dependence, and using sequences to model the temporal dependence. In the domain-oriented services layer, we use graph neural network to embed the graph sequence. We validate the method on an air Spatiotemporal network. Then, we use the embedding to estimate the departure delay time of the flight based on real-time conditions. In the demand-oriented services layer, we design a weighted cross entropy loss function and use a special evaluation to predict the flight departure delay time by the embedding in the domain-oriented services layer. Evaluated through a series of experiments on a real-world data set, we show that the method produces an effective result on the Spatio-temporal network which is substantially better than state-of-the-art alternative task: flight delay estimation. And it performs well in predicting the departure delay time with a total accuracy of 0.87.

Summary The application of anomaly detection to data monitoring is a fundamental requirement of the public service systems of a smart city. Many detection methods have been proposed for identifying anomalous situations, including methods based on periodicity or biseries correlations. However, the detection results of these methods are not ideal. Thus, we present a new anomaly detection algorithm for time series based on the relative outlier distance (ROD) and biseries correlations. The proposed algorithm detects outliers based on the ROD and identifies abnormal points and change points based on biseries correlations. Experimental results show that our method achieves better recall and F1‐measure scores than various time series–based techniques while maintaining a high level of precision.

The information of the real world is stored as triplets (head entity, relation, tail entity) in knowledge graphs. They are extremely useful resources for many intelligent applications but suffer from incompleteness. This paper proposes a knowledge graph representation model to infer latent privacy based on the existing data in attribute network. In our model, considering the nodes are heterogeneous, we classify the nodes into attribute nodes and entity nodes. In order to protect the privacy of entities, we don't follow the previous methods to learn and store the feature embedding of each entity in knowledge graph. Our model focuses in capturing the restriction patterns of attribute nodes, which is safe when merging data from various sources. Given a triplet (entity node, relation, attribute node), firstly, we get the embedding of the entity node by using a sophisticated way to utilize all the information of the node, not only the node connections but also the external text information. Then, we infer the attribute node for the entity node in a certain relation. Finally, we calculate the probability that the triplet is exist. In experiments, we evaluate our model on the tasks of triplet classification and link prediction. Evaluation results show that our approach outperforms the state-of-the-art methods with an accuracy rate of 90.0% in the task of triplet classification on the person attribute knowledge graph FB13. Besides, our model reaches promising performance by MeanRank =5.10, Hits@l = 35.14% and Hits@5=64.94% in the task of conference prediction on the academic network DBLP.

Networks are general tools for modeling numerous information with features and complex relations. Network Embedding aims to learn low-dimension representations for vertexes in the network with rich information including content information and structural information. In recent years, many models based on neural network have been proposed to map the network representations into embedding space whose dimension is much lower than that in original space. However, most of existing methods have the following limitations: 1) they are based on content of nodes in network, failing to measure the structure similarity of nodes; 2) they cannot do well in protecting the privacy of users including the original content information and the structural information. In this paper, we propose a similarity-based graph neural network model, SGNN, which captures the structure information of nodes precisely in node classification tasks. It also takes advantage of the thought of federated learning to hide the original information from different data sources to protect users' privacy. We use deep graph neural network with convolutional layers and dense layers to classify the nodes based on their structures and features. The node classification experiment results on public data sets including Aminer coauthor network, Brazil and Europe flight networks indicate that our proposed model outperforms state-of-the-art models with a higher accuracy.

A serious challenge that confronts shapelet-based algorithms for time series classification is finding optimal shapelets in a short time. Representative shapelet-discovery algorithms find shapelets by evaluating the qualities of candidates extracted from the subsequences. One of the main difficulties is the large amount of time consumed, due to the excessive number of shapelet candidates. To address the above problem, in this paper we propose a fast and interpretable candidate-extraction algorithm to accelerate the process of shapelet discovery. The proposed algorithm utilizes a time series subclass splitting technique to sample time series dataset first. Then, an IDP (Important Data Point)-based selective-extraction strategy is used to extract shapelet candidates. The generated candidates have significant improvements in quality and reductions in quantity. Furthermore, the shapelet candidates generated are more interpretable. To test the effectiveness of the shapelet candidates generated, we transform the original time series and use an off-the-shelf attribute-selection technique to select optimal shapelets from candidates. We then evaluate the proposed algorithm through extensive experiments. The results demonstrate that the proposed algorithm makes significant improvements in accuracy, compared with baselines. Meanwhile, the time consumption is also greatly reduced.

Nowadays, many users select to outsource their job executions to service clouds. These users often have heterogeneous demands while they dynamically arrive at the clouds. For reducing the costs and risks, lots of service cloud operators purchase on-demand instances from public IaaS clouds and provide professional services elastically to users. However, without knowing the future information, it is hard for cloud operators to optimally determine the instance purchasing as well as job scheduling and pricing schemes. In order to achieve maximum social welfare, this paper targets to design an auction mechanism which executes in an online fashion for service clouds, with unique features of job-oriented users, pleasingly parallel jobs and soft deadline constraints. Such a mechanism ought to run in polynomial time, provide truthfulness guarantee, satisfy individual rationality and budget balance, and achieve competitive social welfare. Nevertheless, when designing mechanisms there are a few significant challenges, including the difficulty for finding optimal solution, the strategic behaviours of selfish users with private information and the online arrivals of users. Facing these challenges, we leverage the idea of proportional sharing and propose an online mechanism which is proven to achieve all desired properties. The efficiency of the proposed mechanism is validated by both theoretical analysis and extensive simulations which use both synthetic data and Google's job traces.

In cloud platforms such as Amazon EC2, users can reserve IaaS instances rather than buy on-demand ones to save cost. But it would incur the waste of reservations if there are few demands arriving after reserving instances. Currently, there is a reserved instance marketplace launched by Amazon EC2 cloud, where users can sell their unused instances for avoiding such waste of unused reservations. But for users, it is difficult to make the decision to sell their instances optimally without knowing any information for future demands, for it would incur the extra cost when there are new demands arriving after selling their reservations. For solving this problem, an online selling algorithm is proposed in this paper to guide cloud users in selling reserved instances in Amazon EC2 marketplace. We prove theoretically that our online algorithm Aβ can guarantee a bounded competitive ratio of 2T/β, whose value is specific to the type of reserved instances. Taking the i3.large instance provided by Amazon EC2 as an example, the competitive ratio is 3.36 under its pricing rules for 1-year term. Finally, via extensive experiments using workload data collected from actual applications, we verify our online algorithm's effectiveness and demonstrate that it is much more cost effective to cloud users in IaaS platforms.

Big Data analytics service platform delivers a new type of public cloud offerings, through which end users can outsource their job executions by using a group of professional Big Data processing services in a pay-per-use way. Different from other type of cloud services, parallel jobs dominate the domain of data processing services, whose execution time can be varied greatly with different runtime configurations, such as different degrees of parallelism. In such a market-oriented environment, scheduling jobs from end users efficiently to optimize the Big Data analytics service platform's revenue is a more challenging task. In this paper, we propose a market-oriented heuristic algorithm for scheduling parallel jobs in a Big Data analytics service platform with admission control to optimize the platform operator's revenue. The proposed scheduling heuristic takes into account not only the dynamic revenue gained from accomplishing a job within a specific runtime as well as the consumption of resources needed for running it to achieve this given runtime, but also the potential loss it causes to the system by running this job instead of other waiting jobs currently in the system. We also propose a collaborative filtering based approach to quickly and accurately predict the execution time of parallel jobs running in a Big Data analytics service platform. We have conducted extensive experiments and simulations based on workload data derived from the real-world data analytics service platform and parallel applications. We show that our scheduler can outperform the other scheduling algorithms used for comparison, which are based on classical heuristics from literature, thereby fully evaluating the effectiveness of our market-oriented heuristic scheduling algorithm.

While standard product recommendation systems have proven to be useful for e-commerce, they mainly rely on some prior information about users and products, such as ratings and intrinsic properties of products as well as profile attributes of users. In e-commerce settings, however, a more complete understanding of the demands of customers and the enterprise network constructed by suppliers and manufacturers can be utilized to improve the quality of product recommendations. Moreover, user ratings may be very sparse in some domains. Standard approaches suffer from such data sparsity and neglect to account for important additional dependencies that can be taken into consideration. This motivates us to design a new recommendation model, which incorporates information of network into rating prediction. In this paper, we propose a network-constrained tensor factorization approach, which imposes network constraints as regularization terms on tensor non-negative factorization to improve the accuracy of prediction. To solve the network-constrained regularization problem in our model, we use the Alternating Direction Method of Multipliers (ADMM) method. Experiment results on real-world dataset demonstrate that our approach outperforms other state-of-the-art baselines.

Recently, Amazon EC2 offers a reserved instance marketplace, where cloud users can sell their idle reserved instances varying in contract lengths and pricing options for avoiding the waste of their unused reservations. However, without knowing the future demands, it is hard for users to determine how to sell instances optimally, for it would incur more cost if new demands arrive after selling their reserved instances. For dealing with this problem, in this paper we first propose three online selling algorithms to guide cloud users in making decisions whether or not to sell their reservations in Amazon EC2 marketplace while guaranteeing competitive ratios. We prove theoretically that the three proposed online algorithms can guarantee bounded competitive ratios, whose values are specific to the type of reserved instances under consideration. Specifically, for all standard instances (Linux, US East) for 1-year terms in Amazon EC2, compared with a benchmark optimal offline algorithm, our algorithm A3T/4 can achieve a ratio of 2-α-a/4 in managing instance purchasing cost, where α is the entitled discount due to reservation and a is the selling discount specified by the user who sells its reservations. Finally, through extensive experiments based on workload data collected from real-world applications, we validate the effectiveness of our online instance selling algorithms by showing that it can bring significant cost savings to cloud users compared with always keeping their reservations in Amazon EC2 reserved instance marketplace.

Performance analysis is important for service clouds serving composite service application jobs containing parallelizable tasks, for optimizing the degree of parallelism (DOP) and resource allocation schemes could improve performance obviously. In this paper, we describe a novel tandem queuing network with a parallel multi-station multi-server system as an analytical model for service clouds serving composite service application jobs. We design a partition method (termed the 'pleasing partition') to help us propose an analytical model for parallelizable service which is the vital fraction of composite service. After that, we could obtain a complete probability distribution of response time, waiting time and other important performance metrics calculated by our proposed analytical model. Thus, to use this model, cloud operators could determine proper job configurations and resource allocation schemes, for achieving specific QoS (Quality of Service). Extensive simulations are conducted to validate that our analytical model has high accuracy in predicting performance metrics of composite service application jobs.

In public clouds such as Amazon EC2, there are two main pricing models in purchasing Infrastructure-as-a-Service (IaaS) instances: the pay-as-you-go model and the subscription model. For these two options, users can dynamically combine them to provide services for demands to save their instance acquisition costs. Making optimal decisions toward the purchase of IaaS instances generally requires prior knowledge of future demands; however, it is difficult for users to predict all future workloads accurately. To deal with this problem, online reservation algorithms have been proposed to guide users in reserving instances. However, existing online algorithms do not conform to the pricing rules currently used in public cloud platforms. Therefore, we put forward a new online reserving algorithm for instance in accordance with the pricing policies used in most public IaaS offerings. Specifically, in this study, we use Amazon EC2 as an example to illustrate our algorithm. Through theoretical analysis, we prove that the cost of the proposed algorithm Aβin this paper is not greater than 2-1/β times of the optimal offline algorithm, where β>1 is a critical point in the online reservation algorithm proposed in this paper. Via extensive experimental simulations using both synthetic and actual workload datasets, we demonstrated that the online algorithm Aβis much more cost effective for cloud users than always paying-as-you-go in public IaaS markets.

As more and more users outsource their job executions to service clouds, effective job scheduling and pricing models are needed to solve resource and service competitions between users. Considering the particularity of scheduling and pricing problems in a service cloud whose goal is generally social welfare maximization, current commonly used models, such as fixed-pricing schemes, have obvious shortcomings and thus are unfeasible. Therefore, in this paper, we propose a randomized mechanism to schedule and charge job executions in service clouds. Our proposed mechanism can schedule jobs in a flexible way to achieve approximate social welfare maximization while guaranteeing non-preemption. Flexibility means the number of instances which are allocated to a job can be changed over time. The mechanism is truthful in expectation, computationally efficient and individually rational. The theoretical analysis shows that our mechanism can achieve an expected social welfare approximation ratio α, which can be 2 in some situations. Extensive simulations show that our proposed mechanism can efficiently solve the job scheduling problem in service clouds.

Leveraging Virtual Machine (VM) technologies to host multiple Web applications on the same physical machine can improve the resource utilization and thus save a cloud provider's provisioning cost. By allocating and scheduling virtual CPU (vCPU) resources for running VMs, a hosted Web application may achieve varying performances. Thus, when facing an end user with a specific SLA (Service Level Agreement) requirement, a cloud provider needs to decide how many vCPUs to provision for the target SaaS application to meet the user's end-to-end performance requirement while saving cost. However, it is a non-trivial task to economically determine an optimal resource configuration to meet an end user's SLA requirement. Accurate performance analytic models based on traditional modeling techniques such as queuing systems are difficult to construct for web applications. In this paper, we describe our experience in studying the impact of vCPU provisioning on the performance of 2-tier Web applications, through benchmarking a 2-tier web application in the context of provisioning vCPUs to SaaS applications in a cloud environment. From a cloud provider's perspective, we focus on a generic approach for SaaS benchmarking, which can help to study the impact of vCPU allocations on a Web application's performance and make the optimal vCPU allocation decisions to meet end users' QoS requirements while saving provisioning costs. Besides, based on our benchmark experimental results, we also propose an adaptive controller with a vCPU allocation optimization algorithm which can automatically adjust the vCPU allocations to meet the end users' workload requirements.

MOP (Multiobjective Optimization Problem) is a prevailing research field for its well-modeling on the decision-making dilemma in the real world. We present a distributed ACO (Ant Colony Optimization) algorithm based on a crowdsourcing model, with a few innovative strategies as enhancement, for continuous MOPs. The original MOP is expected to be decomposed into multiple single-objective subtasks, which are then distributed to the individuals on the network, and the non-dominated front for the MOP is constructed by aggregating the solutions from the crowd. Finally, an experiment on the KUR test problem illustrates that our approach is practical.

Parallel application jobs account for a large population in current domain of cloud computing and Big Data processing services, whose execution time can be varied greatly with different runtime configurations. For efficiently scheduling resources and services to run parallel jobs, the ability to quickly and accurately estimate the performance of parallel applications is critical. Analytic predictive models based on traditional modeling techniques such as queuing systems are difficult to construct for parallel applications, due to the high complexity lying in the structures of parallel application models. Furthermore, due to the heterogeneity of resources computing capacities with a scalable computing environment such as a cloud computing platform, performance analytic and prediction becomes increasingly difficult for parallel applications. To address this problem, in this paper we propose a collaborative filtering based approach to quickly and accurately predict the execution time of parallel applications running in heterogenous resources. Particularly, we use the widely used Apache Spark platform as the running framework for parallel applications, and propose a bounds-based performance model to improve the prediction accuracy. Through extensive simulations and experiments on real Spark clusters and two large-scale machine learning applications as well as the simple but classic WordCount sample application, we show that the proposed Collaborative Filtering based approach and bounds-based performance model can accurately estimate the performance of parallel applications.

Mobile cloud computing is now emerging as a promising way to enlarge the capabilities of mobile devices by computation offloading. One of the critical challenges faced by the mobile cloud providers today is how to increase the profitability of their cloud services. In this paper, we deal with the problem of scheduling parallelizable computation jobs offloaded by mobile users in public cloud to maximize cloud providers' profit. We propose an efficient dynamic-priority based and profit-driven scheduling mechanism. We first compute the inherent profitability of a job based on the marginal effect theory and use it as the initial priority. Then, we update the priorities of waiting jobs based on the proposed time-dependent dynamic-priority calculation model as the elapse of time. The results of numerical experiments and simulations show that our approach are efficient in scheduling these kind of jobs in cloud data centers external to mobile devices and considerable profit improvements can be achieved by our proposed dynamic-priority based scheduling mechanism.

Performance analysis is crucial to the successful development of cloud computing paradigm. And it is especially important for a cloud computing center serving parallelizable application jobs, for determining a proper degree of parallelism could reduce the mean service response time and thus improve the performance of cloud computing obviously. In this paper, taking the cloud based rendering service platform as an example application, we propose an approximate analytical model for cloud computing centers serving parallelizable jobs using M/M/c/r queuing systems, by modeling the rendering service platform as a multi-station multi-server system. We solve the proposed analytical model to obtain a complete probability distribution of response time, blocking probability and other important performance metrics for given cloud system settings. Thus this model can guide cloud operators to determine a proper setting, such as the number of servers, the buffer size and the degree of parallelism, for achieving specific performance levels. Through extensive simulations based on both synthetic data and real-world workload traces, we show that our proposed analytical model can provide approximate performance prediction results for cloud computing centers serving parallelizable jobs, even those job arrivals follow different distributions.

QoS aware service composition necessitates an effective pricing mechanism in regulating service providers in public cloud computing environments. However, due to the fact that service providers are usually autonomous, strategic and self-motivated, it is far from trivial to deal with the pricing issues between them. In this paper we formulate a non-cooperative service pricing game to understand the performance of a QoS aware service composition model, for which multiple providers strategically bid how to provide and price their elementary services and establish the Nash equilibrium as the final service composition scheme. We also develop a proportional revenue division rule to incentivize elementary service providers to contribute in improving the QoS of the final composite service delivered to end users. Concerning privacy conservation, we develop a decentralized and recursive bidding algorithm, allowing service providers to reach an equilibrium without disclosing their private information. Through theoretical analysis, we show that a Nash equilibrium exists in a QoS aware service composition game. Through extensive simulations, we show that the proposed recursive bidding process can converge quickly to a Nash equilibrium service composition scheme, and its efficiency is generally high.

Current Infrastructure-as-a-Service (IaaS) clouds offer both on-demand and reservation instance purchasing options. Users can combine these two options dynamically to serve time-varying demands while minimizing their instance acquisition costs. However, when future demands are unknown, it is far from trivial for cloud users to make optimal instance purchasing decisions. To deal with this problem, a carefully designed online algorithm can be employed to guide users in acquiring instances without any prior knowledge of future demands while guaranteeing a competitive ratio. In this paper, we propose an instance reselling model, in which a cloud user can temporarily rent out its idle reserved instances to other users through pay-as-you-go model. We also design online instance acquisition strategies which achieve a better competitive ratio than previous methods. Through extensive simulations based on both synthetic data and real-world traces, we show that our online algorithm under the proposed reselling model can outperform previous models and achieve significant cost savings.

In the case of a cloud-based remote control system such as SCADA (Supervisory Control and Data Acquisition) that enables users to collect data from cloud-connected machines deployed anywhere at any time. However, machine data models may not be updated in a timely manner after the devices are upgrades or modified. This leads to mismatches between the machine data and data models. A key obstacle of the matching is that the machines can be modified. To address this, we present a self-evolving method for machine data model. We give the description of the evolution of machine data models and the self-evolving method for the models in details. The method detects the conflicts between the machine data and models, and transfer or derive models if necessary. Our method can thus facilitate the evolution of machine data models and ensure that every machine in the cloud corresponds to the correct machine data model automatically. At last, we present two case studies to validate our method.

With the development of intelligent manufacturing technology, it can be foreseen that time series data generated by smart devices will raise to an unprecedented level. For time series with high amount, high dimension and renewal speed characteristics, resulting in difficult data mining and presentation on the original time series data. This paper presented a piecewise linear representation based on importance data points for time series data, which called PLR_IDP for short. The method finds importance data points by calculating the fitting error of single point and piecewise, and then represents time series approximately by linear composed of the importance data points. Results from theoretical analysis and experiments show that PLR_IDP reduces the dimensionality, holds the main characteristic with small fitting error of segments and single points.

Today, there has been a strong demand of distributed collaboration in design and manufacturing, due to the acceleration of economic globalization and the popularity of virtual enterprises (VE) model. Because of the characteristics of cloud computing, such as elasticity and on-demand computing, it is promising to deploy and execute collaborative workflows that contain multiple tasks and services such as Computer-Aided Design (CAD) software components on cloud resources for supporting collaboration across enterprises. Specifically, how to cost-effectively select appropriate services to execute workflows within deadlines while without violating multiple constraints becomes an important issue. In this paper, through investigating the practical requirements of collaborative design workflow, we first formulate the issue of the cost-optimal cloud service selection for collaborative workflow executions as a multi-dimensional optimization problem with multiple constraints. Then we propose an effective approach based on genetic algorithms to address this problem for obtaining near-optimal solutions. Based on workload data derived from real-world systems, we conduct experiments which show that our approach outperforms traditional greedy algorithms in finding better solutions and it also provides real-time performance guarantees in real-world cloud computing environments.

A serious challenge that confronts multi-tenant service systems is finding an optimized deployment strategy according to their business scale and operating characteristics. The tenants want to rent high performance services and services providers demand minimizing cost at the same time of meeting the requirements of tenants. But there are often contradictions between high performance and low cost. Therefore, in order to balance the contradiction, this paper proposes a staging-based optimized deployment method. This method performs iterative optimization based on customized workload generation, continuously emulation and evaluation in a benchmark suite. We demonstrate our method by a case study on a multi-tenant Supplier Business Management (SBM) service, as well as evaluate the capability of our benchmark suite through two sets of experiments. Results from these experiments characterized the relationship between workloads and performance, which can help find optimized deployment strategies for multi-tenant applications. In the case study on multi-tenant SBM service system, we gain an optimized strategy that satisfies the requirement of tenants and makes the maximum use of the resources, which can give useful recommendations in real service instances deployment stage.

With the popularity of cloud computing and SaaS, enterprises are willing to rent various cloud services and use them in their daily work. However, on-premise applications are still widely used inside enterprises. That is to say, to achieve the goal of business collaboration, a cloud service usually need to obtain business data from multiple sources, which include the relevant on-premise applications and other cloud services. In this paper, we study and summarize data organization and management patterns for cloud enterprise applications from different aspects. Based on these patterns, we propose an innovative cloud-based data organization model (CEDM). Compared with traditional ones, it highlights the characters of resource sharing and reuse. Besides, this model is more convenient for tackling complex data synchronization issues in cloud environment.

Many automated negotiation models have been developed to solve the conflict in many distributed computational systems. However, the problem of finding win–win outcome in multiattribute negotiation has not been tackled well. To address this issue, based on an evolutionary method of multiobjective optimization, this paper presents a negotiation model that can find win–win solutions of multiple attributes, but needs not to reveal negotiating agents’ private utility functions to their opponents or a third‐party mediator. Moreover, we also equip our agents with a general type of utility functions of interdependent multiattributes, which captures human intuitions well. In addition, we also develop a novel time‐dependent concession strategy model, which can help both sides find a final agreement among a set of win–win ones. Finally, lots of experiments confirm that our negotiation model outperforms the existing models developed recently. And the experiments also show our model is stable and efficient in finding fair win–win outcomes, which is seldom solved in the existing models.