Search

Resource allocation for multi-tenancy application is one key issue for cloud service providers. To deal with the competitive problem of limited resources among tenants, this paper proposes a resource allocation strategy based on game model, which combines tenants’ SLA and their resource utilization, in order to deal with the shortcomings of the traditional allocation strategies that consider few about differentiated resource requirements. A framework for multi-tenant resources allocation based on game is presented, and the corresponding game model is established. The related processes auction and resource allocation are designed. The experimental results indicate that it can meet the personalized and differentiated demand of multi-tenant better, and help to improve the system’s overall resource utilization and decrease the operating costs of cloud service providers.

An effective decentralized metadata management schema is critical to a reliable and scalable large-scale distributed storage system. With regard to the defects of huge expansion cost and high sensitivity to cluster alteration borne in both the hash-based and subtree-based metadata management schemas, a metadata server (MDS) clustering schema based on consistent hash structure, named DSVL, is proposed. By introducing virtual MDS into the consistent MDS cluster, it can effectively balance MDS cluster load. Besides, the standby mechanism and the lazy update policy are merged and applied to the MDS cluster to accomplish rapid MDS failover as well as zero data migration in case of cluster alteration. The results of prototype system and simulation test prove that DSVL, with the features of balanced metadata distribution, rapid failover, flexible scalability and zero data migration amount in case of node alteration, can meet the demands for flexible and effective management of large-scale data storage systems with ever increasing data amount.

Similarity joins on probabilistic data play a vital role in many practical applications, such as sensor reading monitoring and object tracking based on multiple video sources. Earth Mover’s Distance (EMD) proposed in Computer Vision is more effective in returning similar probabilistic data being more consistent to human’s perception to similarity. However, the cubic time complexity of EMD hampers its wide application, especially in the analysis of fast incoming data streams. In this paper we, to the best of our knowledge, make the first attempt to address the EMD similarity join over data streams under sliding window semantics. We first design an efficient and effective index framework, named B + Forests Index, which facilitates data pruning and offers proper strategy to deal with out-of-order data. We then propose the EMD similarity algorithm, named EMD-DSJoin, based on the proposed index framework. We perform extensive experiments on real-world datasets and verify the effectiveness and efficiency of our proposal.

Traditional preferred path mining algorithms cannot accurately identify the level of user interest in a web page, and they are rather complex. In order to solve these problems, the paper proposed an improved calculation method which evaluated the level of user interest in a web page and a mining algorithm of preferred path in the cloud computing environment. Firstly, the paper used Preference Degree to evaluate the user interest in a web page, and then designed a model based on RBF neural network to predict preference degree. This model adequately considered the some key factors such as visit quantity, access time, page bytes, and the nonlinearity relationships among these factors. Secondly, based on preference degree predicting model, the paper presented a new mining algorithm. The algorithm used locations of non-preferred URLs to get the preferred path. The experimental results show that the preference degree prediction model can identify user interest more accurately and comprehensively. In addition, the newly proposed mining algorithm of preferred paths has a lower time complexity and saves more time.

This paper analyzes the effect of Container style middleware on the structure and performance of Component-based system based on architectural patterns, and proposes an approach integrating Container style middleware components and their interaction relation into the application UML (unified modeling language) models. The performance model derived from the integrated UML models can reflect the impact of middleware. So, analysts do not have to know the internal details of middleware at performance modeling. The architectural pattern-based method can be extended to deal with various style middlewares. In the paper, the proposed approach is illustrated by

The first-come first-served scheduling framework adopted by most application servers has been proved to be inappropriate for dealing with unexpected overload for Internet-oriented Web applications. Considering the deficiency of the existing work from the architecture view, this paper presents a scheduling framework based on the notion of QoS benefits, which contains several cooperating components to guarantee the QoS requirements of the applications. The QoS benefits used to evaluate the QoS guarantee provided by the server according to the QoS requirements of the applications and the resource management based on the QoS benefits will help to provide a

Failure detection provides the ability of timely detecting the liveliness of runtime systems and is the basic reliability technology in distributed systems. Providing good failure detection is important to Web application server (WAS) that is the leading middleware in Web computing environment. Adaptive failure detection requires that failure detectors can dynamically adjust the detecting quality according to the requirements of applications and runtime environments. In this paper, the concepts and qualities of failure detectors are firstly discussed, and a multi-level model of failure detection in WAS is presented. Based on the QoS (quality of service) specification of failure detectors, an algorithm for adaptive failure detection is given, and an adaptive failure detection framework of WAS is designed, which can satisfy the requirements of dynamically adjusting qualities and flexible integration of failure detectors. The work has been implemented in OnceAS application server, and the experimental results are