| Recent years, with the popularization of cloud computing and high performance computing, the size and the amount of the datacenter develop rapidly and cause the serious challenges on energy consumption at the same time. According to some reports, the cost of energy consumption accounts for more than the half of operation cost of datacenters. Moreover, the electric energy costed by datacenter is about 1.3% of the total electricity in the world. Therefore, reducing the energy consumption of datacenter is an urgent problem and has become a hot research topic in datacenter.Dynamic Voltage and Frequency Scaling(DVFS) is a wildly applied technique in server for energy saving. DVFS allows the dynamic scaling of voltage and frequency according to current workloads to make the processor work under a more efficient power state to reduce energy consumption. There are many works that focus on the energy-aware scheduling for real-time tasks based on the DVFS, whose objective is to minimize the total energy consumed by a set of tasks in a cluster. Most of these works take the Worst Case Execution Time(WCET) and Deadline as Quality of Service(QoS) requirement, while some tasks’ execution time is not fixed or is hard to be measure. It is important to develop energy-aware scheduling for these tasks to save energy according to their features.A new task model is proposed in this paper, which describes the QoS requirements of tasks from a more common view. This article uses the minimum frequency as the QoS requirement which can describe the tasks with uncertain execution time. Because the system only knows the minimum frequency requirements of tasks without knowing the actual execution time of tasks, it is impossible to estimate the accurate energy con-sumption of tasks under different allocations. Therefore, Energy Consumption Ratio (ECR) is proposed in this paper to describe the relative energy consumption of tasks, which is the ratio of energy consumption under different frequencies to energy under maximum frequency to execute the same task. Based on the proposed definition of new task model and ECR model, this article studies the energy-aware task allocation and scheduling to minimize the total ECR when allocating online tasks for energy saving.By transforming the problem to the Variable Size Bin Packing, we prove that the minimization of ECR is NP-hard in this paper. Because of the difficulty of this problem, we propose task allocation and scheduling method based on the feature of this problem. The proposed methods dispatch the coming tasks to the active servers by using servers as less as possible and adjust the execution frequencies of relative cores to save energy. When tasks finish on servers, there will be some redundant computing resources. If making the migration and frequency scaling for tasks, more energy may be reduced since more cores may work under energy-efficient frequencies. Unfortunately, the task migration among servers usually cause high cost. However, the task migration among cores on the same servers brings little cost, so we apply a task migration mechanism on each individual server to reduce energy. To the best of our knowledge, we are the first to propose ECR to evaluate the energy consumption of tasks under different frequencies and the energy-aware task scheduling combining the task allocation and migration based on the ECR. The experiments in real test-bed system and simulation show that our strategy outperform other strategies compared in this paper, which verify the good performance of our strategy on energy saving. |
PDF Full Text Request