A Data-Driven Based Energy Consumption Modeling for Heterogeneous Servers in Data Centers
Xiaoyan Ma, Changgeng Zhang, Shurong Li, Xinsheng Yang
Abstract
Xiaoyan Ma, Changgeng Zhang, Shurong Li, Xinsheng Yang
Abstract
The server is recognized as one of the main energy consumption equipment in data center, of which accurate energy consumption modeling is the key of improving the data center's energy efficiency. This paper proposes a data-driven based energy consumption modeling for the heterogeneous servers popularized in the modern data center, i.e. it constructs a polynomial based energy consumption model by using the real-time server performance data published in the SPECpower_ssj 2008. The results illustrate that the server can have different energy consumption characteristics even with the same CPU and it also shows the superiority of the proposed model. Lastly, this paper takes the data center with 1000 servers as an example, a comprehensive comparison analysis between the proposed model and the existing server energy consumption models is conduced, and the results show the accuracy of the proposed model and the effectiveness to track the actual energy consumption profile.
OpenAlex reports 4 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
The server is recognized as one of the main energy consumption equipment in data center, of which accurate energy consumption modeling is the key of improving the data center's energy efficiency. This paper proposes a data-driven based energy consumption modeling for the heterogeneous servers popularized in the modern data center, i.e. it constructs a polynomial based energy consumption model by using the real-time server performance data published in the SPECpower_ssj 2008. The results illustrate that the server can have different energy consumption characteristics even with the same CPU and it also shows the superiority of the proposed model. Lastly, this paper takes the data center with 1000 servers as an example, a comprehensive comparison analysis between the proposed model and the existing server energy consumption models is conduced, and the results show the accuracy of the proposed model and the effectiveness to track the actual energy consumption profile.
Key concepts: Server, Energy consumption, Data center, Computer science, Data modeling, Energy (signal processing), Efficient energy use, File server