In this section, presents concerned theories used to deal with the problems arising from resource scheduling in cloud computing. User-level scheduling focuses on resource provision issues between providers and customers, which are solved by economic models. System-level scheduling refers to meta-task execution, sub-optimal solution of which is given by heuristics to speed up the process of finding a good enough answer. Real-time scheduling, which is different from economic and heuristic strategies, is discussed to satisfy the real-time cloud services. Solutions to these pressing needs for cloud environment hinge on VM migration or assignment is designed to move VMs from an overloaded physical machine to a lightly loaded machine. Moving VMs will lessees the burden on the overloaded machine while utilizing the idling physical machine. Moving VMs will reduce the burden on the overloaded physical machine while idle physical machine.
Among the issues is an important parameter threshold that dictates what constitutes a machine underutilized or overloaded. The survey of milojicic 6 identifies the following categories of reason for migration:
Accessing more processing power (in terms of load balancing)
Exploitation of resource locality
Cloud computing is still an evolving paradigm, and it integrates many existing technologies. A brief retrospect of evolution history helps can clarify the situations, opportunities and challenges existing in the development of cloud eco-system. These definitions, attributes, and characteristics will evolve and alter according to the requirements over time. Practically speaking, cloud computing is a service delivery model, where software, platform, infrastructure, information, management can be directly delivered as a service to end consumers. Current efforts are the foundation for further development. After analyzing existing commercial products and research projects, several challenges in terms of middleware, programming model, resource management and business model are highlighted.
Concerned theories including former expressions of problems, algorithms, complexity and schematic methods are briefly introduced. Then scheduling hierarchy in cloud datacenter is presented, by splitting scheduling problem into user-level and system-level. The former focuses on resource provision issues between providers and customers, which are solved by economic models. The latter refers to meta-task execution, a sub-optimal solution of which is given by heuristics to speed up the process of finding a good enough answer. Moreover, real-time scheduling attracts our attention. Most of the work have been done focusing on reduction of allocation cost only. But along with the allocation cost we have to consider whether all the requested virtual machines are getting allocated and whether the service provider is able to serve more request in that particular time-frame
This can be done by allocating a virtual machine to a physical host that best fit the requirements, instead of simply allocating to a host that can serve the request. So an efficient virtual machine placement framework should be developed that can not only allocate the virtual machines using less no of host machines (cost-effective allocation) but also can serve maximum number of requests.