Guanqiu Qi

Two-Tier Multi-tenancy Scaling and Load Balancing

Cloud computing often uses the multi-tenancy architecture where tenants share system software. To support dynamically increasing demands from multi-tenants, the cloud service providers have to duplicate computing resources to cope with the fluctuation of requests from tenants. This is currently handled by virtualization and duplication at the application level in the existing cloud environment, such as Google App Engine. However, duplicating at the application level only may result in significant resource waste as the entire application is duplicated. This paper proposes a two-tier SaaS scaling and scheduling architecture that works at both service and application levels to save resources, and the key idea is to increase the resources to those bottleneck components only. Several duplication strategies are proposed, including lazy duplication and pro-active duplication to achieve better system performance. Additionally, a resource allocation algorithm is proposed in a clustered cloud environment. The experiment results showed that the proposed algorithms can achieve a better resource utilization rate.

DICB: Dynamic Intelligent Customizable Benign Pricing Strategy for Cloud Computing

As cloud services need a fair pricing for both service providers and customers. If the price is too high, the customer may not use it, if the price is too low, service providers have less incentive to develop services. This paper proposes a novel pricing framework for cloud services using game theory (Cournot Duopoly, Cartel, and Stackelberg models) and data mining techniques (clustering and classification, e.g., SVM (Support Vector Machine)) to determine optimal prices for cloud services. The framework is dynamic because the price is determined based on recent usage data and available resources, it is also intelligent as it takes into various economic models into consideration, it is benign because it considers two conflicting parties, service providers and consumers, into consideration at the same time, and it is customizable based on various pricing strategies proposed by service providers and usage patterns as exhibited by consumers. Linear regression is used in various game theory models to determine the optimal price. A global pricing union (GPU) framework is proposed to achieve the best practice of game theory models. Based on the proposed technique, this paper applies this pricing framework to a case study in cloud services, and demonstrates that the prices obtained meet the requirement of traditional supply-demand analysis. In other words, the price obtained is good enough.

Test-Algebra Execution in a Cloud Environment

A algebraic system, Test Algebra (TA), identifies faults in combinatorial testing for SaaS (Software-as-a-Service) applications. SaaS is a software delivery model that involves composition, deployment, and execution of mission application on cloud platforms. Testing SaaS applications is challenging because a large number of configurations needs to be tested. Faulty configurations should be identified and corrected before the delivery of SaaS applications. TA proposes an effective way to reuse existing test results to identify test results of candidate configurations. The TA also defines rules to permit results to be combined, and to identify the faulty interactions. Using the TA, configurations can be tested concurrently on different servers and in any order. This paper proposes one MapReduce design of TA concurrent execution in a cloud environment. The optimization of TA analysis is discussed. The proposed solutions are simulated using Hadoop in a cloud environment.

A novel multi-modality image fusion method based on image decomposition and sparse representation

Abstract Multi-modality image fusion is an effective technique to fuse the complementary information from multi-modality images into an integrated image. The additional information can not only enhance visibility to human eyes, but also mutually complement the limitations of each image. To preserve the structure information and perform the detailed information of source images, a novel image fusion scheme based on image cartoon-texture decomposition and sparse representation is proposed. In proposed image fusion method, source multi-modality images are decomposed into cartoon and texture components. For cartoon components a proper spatial-based method is presented for morphological structure preservation. An energy based fusion rule is used to preserve structure information of each source image. For texture components, a sparse-representation based method is proposed. A dictionary with strong representation ability is trained for the proposed sparse-representation based fusion method. Finally, according to the texture enhancement fusion rule, the fused cartoon and texture components are integrated. The experimentation results have clearly shown that the proposed method outperforms the state-of-art methods, in terms of visual and quantitative evaluations.

Concurrent Test Algebra Execution with Combinatorial Testing

Software-as-a-Service (SaaS), a new software delivery model, plays an important role in daily life. In SaaS, mission-critical applications are composed, deployed, and executed on cloud platforms. SaaS applications needed to have high reliability and availability before publishing. Testing SaaS applications becomes important, as the large number of testing prior to their deployment. Test Algebra (TA), a algebraic system, identifies faults in combinatorial testing for SaaS applications using existing test results and eliminates those related faults. Although TA eliminates a large number of configurations from considerations, it is still difficult to finish testing enormous combinations of services in a reasonable time. To improve TA analysis, this paper proposes a concurrent TA analysis. It allocates workloads into different clusters of computers and performs TA analysis from 2-way to 6-way configurations. Different database designs are used to store the test results of various configurations. Faulty and operational table search algorithms are proposed to retrieve existing test results. One 25-component experiment is simulated using the proposed solutions. The same experiment is also simulated on multiple processors for concurrent TA analysis.

A Cost-Effective Intelligent Configuration Model in Cloud Computing

Cloud computing environments provide a resource pool for customers for their processing, storage and networking needs. It is necessary for customers to choose desirable configuration from vast resources available in the cloud. This paper proposes a heuristic model to choose a cloud configuration for efficient use. Trend analysis in data mining is used to predict the future trend and assist provisioning. The model is experimented with Intel power 32 core machine and in Azure cloud. Both environments demonstrated that the proposed model provided effective solutions.

Integrated fault detection and test algebra for combinatorial testing in TaaS (Testing-as-a-Service)

Abstract Testing-as-a-Service (TaaS) is a software testing service in a cloud that can leverage the computation power provided by the cloud. Specifically, a TaaS can be scaled to large and dynamic workloads, executed in a distributed environment with hundreds of thousands of processors, and these processors may support concurrent and distributed test execution and analysis. This paper proposes a TaaS system based on Adaptive Reasoning ( AR ) and Test Algebra ( TA ) for Combinatorial Testing (CT). AR performs testing and identifies faulty interactions, and TA eliminates related configurations from testing and there can be carried out concurrently. By combining these two, it is possible to perform large CT that were not possible before. Specifically, we performed experiments with 2 50 components with 2.83*10 87 6-way interactions with about 2 1.1 × 10 15 configurations, and this may be the largest CT experimentation as 2014. 98.6% of configurations have been eliminated out of total number of configurations.

Fault-Diagnosis for Reciprocating Compressors Using Big Data

Reciprocating compressors are widely used in the petroleum industry, and a small fault in reciprocating compressors may cause serious issues in operation. Monitoring and detecting potential faults help compressors to continue normal operation. This paper proposes a fault-diagnosis system for compressors using machine-learning techniques to detect potential faults. The system has been evaluated using 100TB operation data collected from China National Offshore Oil Corporation, and the data are first de-noised, coded, and then SVM classification is applied, with 50% of data used for training, the remaining for testing. The results demonstrated that the system can efficiently diagnose potential faults in compressors with 80% accuracy.

Integrated Adaptive Reasoning Testing Framework with Automated Fault Detection

Following development of Software-as-a-Service (SaaS), combinatorial testing technologies are used in SaaS testing. It is difficult to handle the exponential growth of SaaS testing workloads. This paper proposes a new integrated testing framework using Adaptive Reasoning algorithm with automated test cases generation (ARP) and Test Algebra (TA) to increase SaaS testing efficiency. ARP algorithm can rapidly identify and eliminate faulty combinations. The proposed integrated testing framework is simulated in cloud environment.

Autonomous Decentralized Combinatorial Testing

Testing-as-a-Service (TaaS) is a software testing service in a cloud that can leverage the computation power provided by the cloud. Specifically, a TaaS can be scaled to large and dynamic workloads, executed in a distributed environment with hundreds of thousands of processors, and these processors may support concurrent and distributed test execution and analysis. This paper proposes an autonomous decentralized combinatorial testing system based on Adaptive Reasoning (AR) and Test Algebra (TA) for Combinatorial Testing (CT). AR performs testing and identifies faulty interactions, and TA eliminates related configurations from testing and there can be carried out concurrently. By combining these two, it is possible to perform large CT. We performed experiments with 2^10 components and 98:34% of configurations have been eliminated out of total number of configurations by AR and TA analysis.