In order to keep up with the ever-changing world of services, it is essential to fix a broken service composition in a cloud-edge environment when it fails or when users' requests change. However, current methods for replanning or repairing a service composition need to take a while to generate a new solution. To address this issue, we propose using a damaged planning graph repairing approach with graph embeddings to repair broken service compositions in the cloud-edge environment. To begin with, our approach employs graph embedding to acquire the low-dimensional vectors of service nodes within the service network graph. Secondly, we compute the candidate service set by utilizing the failed service, its parameters, and the user's initial input through graph embedding. Thirdly, we design an approach for repairing service compositions in a cloud-edge environment by relying on the candidate service set and the damaged planning graph. To verify the validity of the proposed approach, we analyze and compare the performances of the certain proposed method with referencing re-planning and repair methods, and the proposed method may repair a service composition within a shorter time.

Cloud computing users may obtain great flexibility with low cost by outsourcing their data and services to the cloud. Services are composed together as a solution when no individual services meet the goal. Unfortunately, searching an optimal composition requires significant time due to the high number of available services in the cloud. In this paper, we develop an approach that solves the composition problem with a graph database. This approach uses the shortest bidirectional breadth-first and Dijkstra algorithms to find solutions with either fewest services or the optimal QoS values. Firstly, we preprocess service compositions and store them as paths in a directed bipartite graph in graph database. Compared with using a relational database, the join operator can be avoided and the search speed increases. Secondly, this approach utilize existing resources and can be easily migrated to a cloud database. Correctness and feasibility of this system are verified with experimental results, which shows this system leads to good performance in finding users' satisfying solutions.

The crucial role of networking in cloud computing calls for federated management of both computing and networking resources for end-to-end service provisioning. Application of the Service-Oriented Architecture (SOA) in both cloud computing and networking enables a convergence of network and cloud service provisioning. One of the key challenges to high performance converged network-cloud service provisioning lies in composition of network and cloud services with end-to-end performance guarantee. In this paper, we propose a QoS-aware service composition approach to tackling this challenging issue. We first present a system model for network-cloud service composition and formulate the service composition problem as a variant of Multi-Constrained Optimal Path (MCOP) problem. We then propose an approximation algorithm to solve the problem and give theoretical analysis on properties of the algorithm to show its effectiveness and efficiency for QoS-aware network-cloud service composition. Performance of the proposed algorithm is evaluated through extensive experiments and the obtained results indicate that the proposed method achieves better performance in service composition than the best current MCOP approaches.

Automated service composition fulfills complex tasks by combining different existing web services together. Unfortunately, optimizing service composition is still a challenging area that needs to be addressed. In this article, we propose a novel relational database approach for automated service composition. All possible service combinations are generated beforehand and stored in a relational database. When a user request comes, our system composes SQL queries to search in the database and return the best Quality of Service (QoS) solutions. We test the performance of the proposed system with a web service challenge data set. Our experimental results demonstrate that this system can always find top-K valid solutions to satisfy user's functional and non-functional requirements.

Splitting and merging data is a requirement for many parallel or distributed processing operations. Naive algorithms to split and merge 3D blocks from ultra-high resolution images perform very poorly, as a result of seek times. In contrast, naive algorithms to split and merge 3D slabs perform optimally as seek time is significantly minimized. We introduce and analyze sequential algorithms (Clustered reads, Multiple reads, Clustered writes, and Multiple writes) that leverage memory buffering to address this issue. Clustered reads and Clustered writes, access image chunks only once, but they have to seek in the reconstructed image. Multiple reads and Multiple writes minimize seeks in the reconstructed image, but they access image chunks multiple times. Evaluation on a 3850×3025×3500 brain image shows that our algorithms perform similarly to the optimal configuration provided that enough memory is available. Additionally, Multiple reads supports on-the-fly compression of the merged image transparently but Clustered reads does not, due to its use of negative seeking. We conclude that splitting and merging large 3D images can be done efficiently without relying on complex data formats.

Web service composition enables the provision of existing resources on the web without investing in new infrastructure. However, searching an optimal composition solution with both functional and non-functional requirements is a computationally demanding problem: the time and space requirements may be insufferable due to the high number of available services. To alleviate this problem, we propose the application of a skyline operation to reduce the search space and improve the scalability. We design a system to solve the composition problem with two separate processes. The Graphplan approach finds a solution in a short time, the database approach may take longer time to find a solution, but the solution returned by this approach always has fewer redundant services with a better QoS value. Full Solution Indexing using Database (FSIDB) approach pre-computes all services combinations and store them as paths in a database. Partial pre-composing approach chooses "popular" paths generated by FSIDB approach and store them in a separate table. If the problem can be solved by these paths, there is no need to search the table with whole paths. We evaluate our approach with a web service challenge dataset.

Objective: To analyze the strain distribution and drug sensitivity of Candida mycoderma isolated from patients with vulvovaginal candidiasis(VVC), to provide evidence for clinical rational drug use. Method:119 clinical strains were from patients' leucorrhea sample. Those patients saw a doctor in TCM hospital of Guangdong Province's Gynaecological and Dermatological department. Those clinical strains were detected by Chromagar candida colored culture medium and API / ATB kits, fungus drug sensitivity slab. Result: When it came to strain distribution, candida albicans accounted for 81.51%, Candida glabrata accounted for 15.13%, Candida tropicalis accounted for 1.68%, Candida parapsilosis accounted for 1.68%. When it confered to susceptibility test, the susceptibility of strains to amphotericin B was strongest. Conclusion: Candida albicans is the most important pathogenic bacterium of VVC. Other kinds of Candida mycoderma are more and more. Widespread usage of fluconazol is responsible for this phenomenon. In clinic, we should pay more attention to susceptibility test and rational use of drug to elevate therapeutic efficacy.

Psoriasis is a common chronic skin disease of which the pathogenesis has not yet been fully elucidated. Traditional Chinese medicine (TCM) has effectiveness in treating this disease. Syndromes are the basic pathological units and key concepts of TCM theory, standardisation in TCM is significantly important for psoriasis diagnosis and treatment. Nevertheless, there are few documents concerning this issue. This study focused on psoriasis vulgaris, collected and analysed the literature in recent decades (1979-2011), applied data mining methods to analyse the characteristics of the main type symptoms, and formed psoriasis vulgaris syndrome diagnostic criteria.

QoS-aware service composition is the generation of a business process to fulfill functional goals and optimize the QoS criteria at the same time. People may focus on the optimization of a single QoS criterion or a set of QoS criteria. We find that though many composition algorithms can get the optimal QoS values, the solutions obtained can possibly contain redundant services, the removal of which does not worsen the QoS value of the solution. In the literatures using Web Service Challenge (WSC) open data sets, the removable services can be over 30% of the services in the solutions. This common problem has been ignored so far. The fundamental reason is that execution cost is not one of the criteria to optimize. Even in the cases when execution costs are not explicitly given for each service, we are still motivated to reduce the number of services in the final solution by assuming each service takes unit cost. In this paper, we study the redundancy removal problem to further optimize the QoS optimal solutions obtained by QoS-aware service composition algorithms. We model the redundancy removal problem as an integer programming problem. Though solvable using a standard solver, we present an algorithm to solve the problem in this specific context and it proves to have better performance than a standard integer programming solver. We also present the results of our data experiments.

The crucial role that networking plays in Cloud computing calls for a holistic vision that allows combined control, management, and optimization of both networking and computing resources in a Cloud environment, which leads to a convergence of networking and Cloud computing. Network virtualization is being adopted in both telecommunications and the Internet as a key attribute for the next generation networking. Virtualization, as a potential enabler of profound changes in both communications and computing domains, is expected to bridge the gap between these two fields. Service-Oriented Architecture (SOA), when applied in network virtualization, enables a Network-as-a-Service (NaaS) paradigm that may greatly facilitate the convergence of networking and Cloud computing. Recently the application of SOA in network virtualization has attracted extensive interest from both academia and industry. Although numerous relevant research works have been published, they are currently scattered across multiple fields in the literature, including telecommunications, computer networking, Web services, and Cloud computing. In this article we present a comprehensive survey on the latest developments in service-oriented network virtualization for supporting Cloud computing, particularly from a perspective of network and Cloud convergence through NaaS. Specifically, we first introduce the SOA principle and review recent research progress on applying SOA to support network virtualization in both telecommunications and the Internet. Then we present a framework of network-Cloud convergence based on service-oriented network virtualization and give a survey on key technologies for realizing NaaS, mainly focusing on state of the art of network service description, discovery, and composition. We also discuss the challenges brought in by network-Cloud convergence to these technologies and research opportunities available in these areas, with a hope to arouse the research community's interest in this emerging interdisciplinary field.

When two Web services work together, they exchange messages in a predefined interface process. Two interface processes should be compatible when they can work properly. Our idea to fix incompatibility problem in service processes is to change an incompatible process so that the new process can simulate a compatible process. We consider not only the control flow but also the data flow in modeling the processes into FSMs. We present a technique that not only detects the incompatibility, but also provides resolution strategies to generate the new process.

Service composition is nowadays mainly seen as a once-for-all activity. Supporting a dynamic service world, where both available services and needs may change, requires runtime adaptive features for service composition. In this paper we propose a repair technique for internal composition adaptation, as opposed to external adaptation. Moreover, setting up our proposal in the planning framework, we compare our repair technique with reference to re-composition, that is re-planning.

In machine learning and data mining, traditional learning models aim for high classification accuracy. However, accurate class probability prediction is more desirable than classification accuracy in many practical applications, such as medical diagnosis. Although it is known that decision trees can be adapted to be class probability estimators in a variety of approaches, and the resulting models are uniformly called Probability Estimation Trees (PETs), the performances of these PETs in class probability estimation, have not yet been investigated. We begin our research by empirically studying PETs in terms of class probability estimation, measured by Log Conditional Likelihood (LCL). We also compare a PET called C4.4 with other representative models, including Naïve Bayes, Naïve Bayes Tree, Bayesian Network, KNN and SVM, in LCL. From our experiments, we draw several valuable conclusions. First, among various tree-based models, C4.4 is the best in yielding precise class probability prediction measured by LCL. We provide an explanation for this and reveal the nature of LCL. Second, compared with non tree-based models, C4.4 also performs best. Finally, LCL does not dominate another well-established relevant metric — AUC, which suggests that different decision-tree learning models should be used for different objectives. Our experiments are conducted on the basis of 36 UCI sample sets. We run all the models within a machine learning platform — Weka. We also explore an approach to improve the class probability estimation of Naïve Bayes Tree. We propose a greedy and recursive learning algorithm, where at each step, LCL is used as the scoring function to expand the decision tree. The algorithm uses Naïve Bayes created at leaves to estimate class probabilities of test samples. The whole tree encodes the posterior class probability in its structure. One benefit of improving class probability estimation is that both classification accuracy and AUC can be possibly scaled up. We call the new model LCL Tree (LCLT). Our experiments on 33 UCI sample sets show that LCLT outperforms all state-of-the-art learning models, such as Naïve Bayes Tree, significantly in accurate class probability prediction measured by LCL, as well as in classification accuracy and AUC.

Diagnosability is a crucial property that determines at design stage how accurate any diagnosis algorithm can be on a partially observable system. Recent work on diagnosability has generalized fault event case to pattern case, which can describe more general objectives for diagnosis problem, but based on global model and global twin plant construction. In this paper, we propose an original framework to solve pattern diagnosability in a distributed way to avoid calculating global objects. We first show how to incrementally accomplish pattern recognition without building global model by propagating only diagnosability relative information between components. Then an efficient way to construct pattern verifier is proposed, which is inspired from the classical twin plant method but with smaller state space, to search for partial critical paths, whose global consistency is subsequently checked. Meanwhile we prove that the result obtained from our distributed approach is on an equality with that from the centralized one but the evaluation result shows that our search state space exploited is only a small subpart of the global twin plant, whose construction is unavoidable in the centralized approach.

This paper studies the automatic scheduling problem at the Canadian national synchrotron facility, Canadian Light Source (CLS). An automatic scheduling tool needs to be developed to replace the current manual approach for scheduling experiments on a set of beamlines - resources that generate high-intensity X-rays for use in many kinds of scientific experiments. We present an, Integer programming model for this scheduling activity by formulating it as a problem of unrelated and paralleled machines with partially overlapping capabilities. Furthermore a heuristic based approach is used that can save computation time by pruning the search space. Using realistic data sets generated using parameters made available by CLS, we compare the performance of the base line approach that uses ILOG CPLEX implementation of the Integer programming algorithm with one that uses heuristics. The results show that the heuristic approach runs faster than the base-line, but at the cost of producing a less optimal scheduling solution. An obvious advantage of the study presented in this paper is that the automatic scheduling can handle more scheduling conditions and constraints than humans are able to handle manually and can reach optimal solutions. As far as we know, this is the first attempt to propose an automatic scheduling approach for synchrotron facilities like CLS around the world.

In this paper, we have studied a common Web service composition problem, the syntactic matching problem, where the output parameters of a Web service can be used as the input parameters of another Web service. Many automatic Web service composition algorithms based on AI planning techniques have been proposed. However, most of them do not scale well when the number of Web services increases, or may miss finding a solution even if one exists. The planning graph, another AI planning technique, provides a unique search space. We have found that when we model the Web service composition problem as a planning graph, it actually provides a trivial solution to the problem. Instead of following the usual way to find a solution by a backward search, we put our efforts into removing the redundant Web services contained in the planning graph. Our approach can find a solution in polynomial time, but with possible redundant Web services. We have tested our algorithms on the data set used in ICEBEpsila05 and compared our results with existing methods.

Web service orchestration languages describe executable business processes composed of Web services. A business process can fail for many reasons, such as faulty Web services or mismatching messages. It is important to find out which Web services are responsible for a failed business process because we could penalize these Web services and exclude them from the business process in the future. In this paper, we propose a model-based approach to diagnose orchestrated Web service process. We convert the Web service orchestration language, BPEL4WS, into synchronized automata, so that we have a formal description of the topology and variable dependency of the business process. After an exception is thrown, the diagnoser can calculate the business process execution trajectory based on the formal model and the observed evolution of the business process. The faulty Web services are deduced from the variable dependency on the execution trajectory. We demonstrate our diagnosis technique with an example.

Existing work shows that classic decision trees have inherent deficiencies in obtaining a good probability-based ranking (e.g. AUC). This paper aims to improve the ranking performance under decision-tree paradigms by presenting two new models. The intuition behind our work is that probability-based ranking is a relative metric among samples, therefore, distinct probabilities are crucial for accurate ranking. The first model, lazy distance-based tree (LDTree), uses a lazy learner at each leaf to explicitly distinguish the different contributions of leaf samples when estimating the probabilities for an unlabeled sample. The second model, eager distance-based tree (EDTree), improves LDTree by changing it into an eager algorithm. In both models, each unlabeled sample is assigned a set of unique probabilities of class membership instead of a set of uniformed ones, which gives finer resolution to differentiate samples and leads to the improvement of ranking. On 34 UCI sample sets, experiments verify that our models greatly outperform C4.5, C4.4 and other standard smoothing methods designed for better ranking

Accurate probability estimation generated by learning models is desirable in some practical applications, such as medical diagnosis. In this paper, we empirically study traditional decision-tree learning models and their variants in terms of probability estimation, measured by conditional log likelihood (CLL). Furthermore, we also compare decision tree learning with other kinds of representative learning: Naive Bayes, Naive Bayes tree, Bayesian network, K-nearest neighbors and support vector machine with respect to probability estimation. From our experiments, we have several interesting observations. First, among various decision-tree learning models, C4.4 is the best in yielding precise probability estimation measured by CLL, although its performance is not good in terms of other evaluation criteria, such as accuracy and ranking. We provide an explanation for this and reveal the nature of CLL. Second, compared with other popular models, C4.4 achieves the best CLL. Finally, CLL does not dominate another well-established relevant measurement AUC (the area under the curve of receiver operating characteristics), which suggests that different decision-tree learning models should be used for different objectives. Our experiments are conducted on the basis of 36 UCI sample sets that cover a wide range of domains and data characteristics. We run all the models within a machine learning platform $Weka

Composing business processes from individual services can be viewed as a planning problem in which a planner determines the execution orders of services in a process. Most existing Web service composition research considers connecting Web services into a business process. We argue that most existing Web services are informative Web services that are not the actual business services, but give the parameters of their correspondent business services. The planning problem is not only to select the proper business services, but also to determine the parameters of the business services which affect the ordering of the business services. Furthermore, it is not possible to extract all information from informative Web services through queries. The planner has to work with the problem space that is not fully enumerable. This paper presents a method to optimize planning results with incompletely observed problem space. Genetic algorithms (GA) help to navigate the incompletely observed problem space. At each loop of GA, Web service data are queried and a new sub problem space is built. The planner works with the sub problem space and calculates all feasible plans. The plans are evaluated by GA in fitness function and the best plans are kept for the next loop of GA. The fitness function of GA reflects domain-dependent user preferences. The selected final plan is an optimized feasible plan though global optimization is not guaranteed

Online experimentation is one of the promising e-learning applications that the students operate the remote instruments. It is well motivated to use Web services to architect this kind of applications due to the advanced ability of Web services to bridge different platforms and languages. We present our work in using Web services for prototyping a service-oriented architecture for online experiment systems in this paper. We present the methodology to wrap the functions of instruments into Web services and the advanced features the instrumental Web services should have.

Monotonicity is a simple yet significant qualitative characteristic. We consider the problem of segmenting an array in up to K segments. We want segments to be as monotonic as possible and to alternate signs. We propose a quality metric for this problem, present an optimal linear time algorithm based on novel formalism, and compare experimentally its performance to a linear time top-down regression algorithm. We show that our algorithm is faster and more accurate. Applications include pattern recognition and qualitative modeling.

Online experimentation allows students from anywhere to operate remote instruments at any time. This promising e-learning application is well positioned to use Web services to conduct online experiment systems due to its interoperability and Internet compliance. We present a double client-server architecture for online experiment systems and the methodology to wrap the functions of instruments into Web services. We propose that the instrument Web services should be stateful services and we present the framework to manage the states of the instrument Web services. We benchmark the performance of this system when using SOAP as the wire format for communication and propose solutions to optimize performance.

Workflow management techniques are aimed at supporting business processes across organizational boundaries. Current techniques lack the formal tools to model and analyze workflows on a large scale, and current WFMSs do not have the ability to react to the response of another WFMS dynamically. Using Petri nets as the modeling tool, we present the concept of a standard workflow structure, and a set of standard workflow blocks are designed. We prove that modeling a workflow process in standard workflow structures can guarantee the soundness of a workflow network as defined by W.M.P. van der Aalst (1998). We also present a method of decomposing a large-scale workflow network into sub-networks. From our analysis and proof, we draw the very good conclusion that a workflow network composed of self-loop-connected sub-networks maintains soundness under proper structural conditions. The method presented is not only a design language for the specification of complex workflows, but also a powerful analysis technique to verify the workflow procedures.

Both workflow and agent technology have recently been applied to business process management. The integration of these two technologies definitely provides solutions to problems that cannot be solved by either of them individually. This paper summarises the capabilities of these two technologies and discusses the forms and benefits of integrating them for business process management. Generally, agent-enhanced workflow management and agent-based workflow management are the main forms of application of intelligent agents to workflow systems. Some research issues in each form are discussed. A conclusion, with a discussion about future research directions, is also given.