The Study On Instance-Intensive Workflow Scheduling Algorithms In Community Cloud

The problem of solving fast collaboration between many related enterprises has always been an important research hotpot in computer network research. At present, the rapid development of network technology is in dire need of a new interconnected network structure which can provide a unified coordination mechanism. In recent years, the putting forward of cloud computing concept and related frameworks has provided a good solution. As a typical cloud computing application pattern, the community cloud pattern and related systems are particularly prominent in this area.Community cloud is a resources community formed by services, hardware resources and other network resources from enterprises and institutions. Through making flexible integration in community cloud organizations, the resources in side of the community cloud can achieve the final aim of fast collaboration.The goal of this thesis is to realize a workflow based fast collaboration mechanism for community cloud pattern, and propose a community cloud workflow oriented scheduling strategy which can provide good task scheduling solutions in the community cloud workflow systems, and solve the problem of instance-intensive scheduling brought about by high concurrent requests of cloud applications. The works of this thesis are listed below:First, this thesis proposes a community cloud workflow system framework, and describes the working principles of it. The community cloud workflow system framework is composed by three major part:service node, management node and service catalog, in which service node is the abstraction of network resources for performing community cloud member services and the ultimate executer of collaborative services, management node is responsible for coordinating the execution of the entire community, providing task allocation and scheduling for service nodes, and service catalog keep all work state and QoS information of every service node of community cloud members, providing service retrieval for the creation of cooperative domains. Composed by these structures, the community cloud workflow system framework is able to fit the community cloud resource management model, and meet the need of fast collaboration of community cloud well.Second, after intensive studies in distributed workflow scheduling algorithms, this thesis proposes an Unified Scheduling Strategy for Instance-intensive workflows USS-I. Through the method of global QoS decomposition, scheduling between cloud members and scheduling inside cloud members, the strategy can make an unified workflow scheduling for the instance-intensive workflow applications, realize real-time load balancing of resources, and provide good support for the fast collaboration between community cloud members.Third, in this thesis the three algorithms of USS-I are implemented, and the contrast experiment of scheduling algorithm in side cloud member QoS-constrained Pre-calculate Scheduling Algorithm (QCPCSA) is made, particularly. It shows that the algorithm can improve system performance when the factor x is set in a specified range. The integrated results show that the USS-I strategy can meet the needs of the community cloud workflow system very well.The outcome of this thesis is core technology of large-scale network applications in community cloud application pattern, it has very important meaning for the improvement of the workflow execute efficiency, increase of the throughput of concurrent process instances, and improvement of the load balancing of resources. Meanwhile, the outcome of this thesis is also one of the core technology of workflow and service centralized infrastructure in cloud environment, it must have a good prospect of application and extension in the near future.