研究目的
To optimize solar energy consumption in distributed clouds by proposing a stochastic modeling approach named SAGITTA, which aims to minimize renewable energy loss by adequately allocating computing resources to incoming user requests.
研究成果
SAGITTA demonstrates good results in terms of brown energy consumption, with a difference of only 5.2% with the optimal solution. It outperforms traditional approaches in all cases and can smoothly scale with the number of data centers belonging to the cloud. Future work includes extending SAGITTA by integrating the ability to dynamically migrate virtual machines and the energy production from one site to another, adapting it to continuous workloads, and integrating the impact of network devices on the energy consumption.
研究不足
The study does not explicitly model the brown power production, assumes the telecommunication network to have negligible impact on the system functioning, and does not take into account the energy consumed by the data centers’ cooling systems. Additionally, the dynamic programming algorithm for computing the optimal energy consumption requires significant computing resources.
1：Experimental Design and Method Selection:
The study employs a stochastic modeling approach for estimating renewable energy production and greedy heuristics for allocating resources to incoming user requests and switching off unused servers.
2：Sample Selection and Data Sources:
Real recordings of green power production and real workload traces are used. The green power production data comes from the Photovolta project of the University of Nantes, and the workload input is based on the normalized ClarkNet HTTP trace.
3：List of Experimental Equipment and Materials:
The simulation involves data centers composed of clusters with homogeneous nodes based on the Taurus servers of the French experimental testbed Grid’
4：Experimental Procedures and Operational Workflow:
50 The controller implementing the SAGITTA approach is run at each time slot (every five minutes), saving all the data received from the green power sources during the current day and computing the standard deviations using this history.
5：Data Analysis Methods:
The performance of SAGITTA is compared against two Round-Robin inspired algorithms through simulation-based evaluation.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
