Transactions on Large-Scale Data- and Knowledge-Centered Systems XXXI

This paper deals with models and tools allowing designers of geographic web application to focus on the design work rather than on the related coding problems. The contributions of this paper are a specific design model and its associated design environement. The proposed design model is composed of elements that can be translated, through transformation model technics, into executable source code taking into account high-level specifications of the designers. This automated code generation property offers a design approach based on short cycles where designers may adjust their specifications until the generated application matches their requirements. To facilitate this process, the proposed model has been integrated into a graphical web-based environment allowing designers to visually express their specifications and then to generate and to execute the specified application.

Monitoring events in communication and computing systems becomes more and more challenging due to the increasing complexity and diversity of these systems. Several supporting tools have been created to assist system administrators in monitoring an enormous number of events daily. The main function of these tools is to filter as many as possible events and present highly suspected events to the administrators for fault analysis, detection and report. While these suspected events appear regularly on large and complex systems, such as cloud computing systems, analyzing them consumes much time and effort. In this study, we propose an approach for evaluating the severity level of events using a classification decision tree. The approach exploits existing fault datasets and features, such as bug reports and log events to construct a decision tree that can be used to classify the severity level of other events. The administrators refer to the result of classification to determine proper actions for the suspected events with a high severity level. We have implemented and experimented the approach for various bug report and log event datasets. The experimental results reveal that the accuracy of classifying severity levels by using the decision trees is above 80%, and some detailed analyses are also provided.

Kerberos is a distributed authentication protocol which guarantees the mutual authentication between client and server over an insecure network. After the identification, all the subsequent communications are encrypted by session keys to ensure privacy and data integrity. Nowadays, many traditional authentication systems have tried moved to biometric system for convenience. However, the security and privacy of these systems need to put on the table. In this paper, we have proposed an efficient hybrid approach for protecting biometrics in remote authentication protocol based on Kerberos scheme. This protocol is not only resistant against attacks on the insecure network such as man-in-the-middle attack, replay attack,… but also able to protect the biometrics for using fuzzy extractor and non-invertible transformation. These techniques conceal the user’s cancelable biometrics into the cryptographic key called biometric key. This key is used to verify a user in authentication phase. Therefore, there is no need to store users’ plaint biometrics in the database. Even if biometric key is revealed, it is impossible for an attack to infer the users’ biometrics for the high security of the fuzzy extractor scheme. Moreover, another remarkable contribution of this work is that a user can also change his biometric key without replacing his biometrics. The protocol supports multi-factor authentication to enhance security of the entire system.

We propose new parallel learning algorithms of local support vector machines (local SVMs) for effectively non-linear classification of large datasets. The algorithms of local SVMs perform the training task of large datasets with two main steps. The first one is to partition the full dataset into k subsets of data, and then the second one is to learn non-linear SVMs from k subsets to locally classify them in parallel way on multi-core computers. The k local SVMs algorithm (kSVM) uses kmeans clustering algorithm to partition the data into k clusters, then constructs in parallel non-linear SVM models to classify data clusters locally. The decision tree with labeling support vector machines (tSVM) uses C4.5 decision tree algorithm to split the full dataset into terminal-nodes, and then it learns in parallel local SVM models for classifying impurity terminal-nodes with mixture of labels. The krSVM algorithm is to train random ensemble of kSVM. The numerical test results on 4 datasets from UCI repository, 3 benchmarks of handwritten letters recognition and a color image collection of one-thousand small objects show that our proposed algorithms of local SVMs (kSVM, tSVM, krSVM) are efficient compared to the standard SVM (LibSVM) in terms of training time and accuracy for dealing with large datasets.

Business services arguably play a central role in service-based information systems as they would fill in the gap between the technicality of Service-Oriented Architecture and the business aspects captured in Enterprise Architecture. Business services have distinctive features that are not typically observed in Web services, e.g. significant portions of the functionality of business services might be executed in a human-mediated fashion. The representation of business services requires that we view human activity and human-mediated functionality through the lens of computing and systems engineering. Contractually specifying a business service is crucial for the design and operationalization of business services from the service providers’ point of view. In this article, we present an overarching modeling and formalization approach to the contractual specifications of business services. First, business services are conceptually described from three different perspectives, giving rise to a list of service descriptors that matter most for the contractual specifications of services. Second, we formalize the service descriptors. Third, we devise a formal machinery to (a) verify if a group of services contractually match the specification of the bulkier service in question; (b) to assess the contractual proximity of service groups relative to a contractual service specification to help decide which combination of services from a catalog best realize the desired functionality.

Energy efficiency has become an important measurement of scheduling algorithms for Infrastructure-as-a-Service (IaaS) clouds. This paper investigates the energy-efficient virtual machine scheduling problems in IaaS clouds where users request multiple resources in fixed intervals and non-preemption for processing their virtual machines (VMs) and physical machines have bounded capacity resources. Many previous works are based on migration techniques to move on-line VMs from low utilization hosts and turn these hosts off to reduce energy consumption. However, the techniques for migration of VMs could not use in our case. The scheduling problem is NP-hard. Instead of minimizing the number used physical machines, we propose a scheduling algorithm EMinTRE-LDTF to minimize the sum of total busy time of all physical machines that is equivalent to minimize total energy consumption. In this paper, we present the proved approximation in general and special cases of the scheduling problem. Using Feitelson’s and Lublin99’s parallel workload models in the Parallel Workloads Archive, our simulation results show that algorithm EMinTRE-LDTF could reduce the total energy consumption compared with state-of-the-art algorithms including Tian’s Modified First-Fit Decreasing Earliest, Beloglazov’s Power-Aware Best-Fit Decreasing and Vector Bin-Packing Norm-based Greedy. Moreover, the EMinTRE-LDTF has less total energy consumption compared with our previous heuristic (e.g. MinDFT) in the simulations.