Font Size: a A A

Research On Dependency Aware Construction Method Of Multi-Tenant Application

Posted on:2012-08-05Degree:DoctorType:Dissertation
Country:ChinaCandidate:Y ZhangFull Text:PDF
GTID:1228330371950978Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
With the popularization of Internet and rapid development of software technology and application pattern, software is showing its characteristics of network, platform and service. Software as a Service is a new service delivery model, which has the advantage of on-demand using, no maintenance cost and easy to extend. SaaS is changing the way people using software and the way ISV providing software. On the other hand, with the rapid development of distributed computing, parallel computing, grid computing, cloud computing and the popularization of virtualization and utility computing, SaaS is facing many challenges and chances.Analysis with the theory of supply and demand in Modern economics, the establishments of any business model needs the support of a kind of relationship between supply and demand. In traditional business model, software is a kind of product. However, software is delivered as service in SaaS model. As a result, SaaS model is an important and special business model in modern service industry.On the other hand, the success of SaaS business model needs the support of well designed IT architecture. Multi-tenancy is one of the core technologies to realize SaaS model, which can host a large number of tenants on the same software and hardware infrastructure. With Multi-tenancy technology, each tenant can use the resources on-demand, and customize the software without impact on other tenants. It can reduce the cost of service delivery, maintenance and get more income with economy of scale. Therefore, Multi-tenancy is attractive in both academe and industry.At present, in the research of Multi-tenancy, there are several problems need to be addressed:service-oriented modeling, customization, tenant placement and performance guarantee mechanism, etc. Aimed at these problems and with the support of National High Technology Research and Development Program ("863 "Program) and The National Natural Science Foundation of China (NSFC), this paper focuses on the key technical problems in the construction method of dependency-aware service oriented multi-tenancy application.The main contributions of the thesis are as follows:1. Proposed Dependency Aware Hierarchical Service Model (DAHSM)Based on the relationship between services, this paper extends the traditional service registration and management mechanism by annotation of service relationship. According to the requirement of rapid and flexible construction of SaaS, this paper partitions SaaS with level and granularity attributes. This paper proposes the DAHSM model and uses extended hyper-graph to formalize it, and maps the construction of SaaS to the establishment and instantiation of DAHSM. Also, this paper gives verify and dependency computing API to serve other process, such as service selection and tenant placement, etc.2. Proposed performance customization strategy based on dynamic service selection.In order to address the problem of performance customization, this paper proposes dynamic service selection based customization strategy. Under the scene that a logic service may have many instances with different QoS attributes, this paper transforms the problem of performance customization into the problem of SLA driven dynamic service selection. First, partition all tenants into several set with clustering algorithm. Then, according to the highest SLA in each set, execute Genetic Algorithm to select suitable service instance for each logic service to make the total QoS can meet the SLA of each tenant. The service selection results should be dynamically binding to concrete service instance.3. Proposed tenant placement strategy and algorithm based on the resource consumption estimation modelIn order to address tenant placement problem and reuse of services, this paper defines function of similarity to match objective service by both functional and non-functional attributes to compute the similarity between tenants’ requirement and objective services. Meanwhile, this paper defines a resource consumption estimation model to predict the resource consumption after tenants are placed, and check whether SLA could be met. When there are several services can be used, the resource utilization should be considered.4. Proposed load-based optimized service deployment strategy and algorithmThis paper researches load-based performance guarantee mechanism with constrains of limited resources. This paper defines load balanced partition, and proposed two step service deploy algorithm named first deployment and reallocation to get load balance. Meanwhile, this paper proposes Markov chain based business predict mechanism to deal with peak load or Unexpected situations. Meanwhile, with the help of DAHMS model, the hot spot services could be detected and processed.On the basis of above research, this paper develops a prototype system to verify the related work under the background of supplier relationship management (SBM) service. This paper then intruduces future work on DAHSM model, tenant placement problem and performance guarantee mechanism.
Keywords/Search Tags:Multi-tenancy, Dependency-aware, Hypergraph, Service selection, Tenant placement
PDF Full Text Request
Related items