Research On Energy-Aware Scheduling Technology For Cloud Platform

With arriving of big data times, cloud computing has aroused general concern of industry and academia. This requires that the cloud computing service providers extensively deploy large-scale infrastructure to provide many kinds of resource. Infrastructure as a Service (IaaS), as implemented by Amazon EC2 and Microsoft Azure, has quickly moved into our working and life. IaaS providers supply automatic and flexible resources to customers. Cloud computing technology provides a convenient to us. But at the same time, these datacenters result in high energy consumption problem. The efficient management of energy consumption becomes a research topic which is concerned by scholars in recent years.There are many kinds of tasks in cloud datacenter. If we allocate these resource reasonable, it will save energy consumption. Resource allocating to these tasks falls into the field of multi-objective optimization. We build the objective function and the variable factors through the system model. Then we put forward an energy-aware scheduling algorithm—Task Set Consolidation Algorithm (TSC). TSC can control the task consolidation by comparing the dist of each task set. Our simulation experiment compares TSC with other algorithms. The result shows that TSC is about 67% greater than no consolidation in the active physical server number, as well as about 70% greater than no consolidation in the energy consumption.A growing segment of data-intensive workloads are managed using MapReduce-style platforms. With the popularity of virtualization technology, MapReduce virtual cluster gradually becomes the service provided directly by IaaS clouds. We study batch-oriented placement and online placement for virtual machine placement energy-aware problem, respectively. For batch cases, we propose Virtual Machine Tight Recipe Placement Algorithm and Frequency Scaling Algorithm to save energy while guarantee job SLAs. We prove the most efficient frequency that minimizes the energy consumption, and the upper bound of energy saving through DVFS techniques. FS can also be used in combination with other placement algorithms. For online case, we propose Online Time Balance Algorithm. OTB can balance the server duration and utilization and save energy.Finally, we implement the visualization system -- Cloud Computing Scheduling Management System. Our System can quickly deploy Hadoop platform, and configure in personal demand. Users are enable to configure cluster on graphical configuration interface by using our system. At the same time, we provide a variety of optional schedulers which are implemented by our project team. User can no longer be constrained by the Hadoop platform’s default scheduler, and choose suitable scheduler. Our system can monitor the running status of Hadoop platform real time.