The behavior of a distributed control algorithm in a data center to optimize power consumption

Biel, M.R. (2017) The behavior of a distributed control algorithm in a data center to optimize power consumption. Bachelor's Thesis, Industrial Engineering and Management.

	Text Bachelor_IEM__2017_MarkBiel.pdf - Published Version Restricted to RUG campus only Download (2MB)
	Text toestemming.pdf - Other Restricted to Backend only Download (80kB)

Abstract

To optimize power consumption in data centers, thermal aware job scheduling is utilized. Thermal aware job scheduling takes into account the complex heat flows in data centers to migrate workload in order to prevent local hotspots. By preventing local hotspots from arising, the cooling equipment cools more efficient and thus optimizes power consumption. This research explores a distributed thermal aware job scheduling algorithm. Taking into account the increasing size of data centers, this approach yields several advantages in terms of implementation and applicability. While the distributed control algorithm discussed in this research has been evaluated within the proposed framework, it is unknown how the algorithm copes with a more realistic situation. Therefore, the distributed control algorithm is evaluated through MATLAB simulations while subject to different network conditions and constraints. Results from the simulations are utilized to draw conclusions regarding the behavior of the algorithm in terms of four Key Performance Indicators. The results show that the algorithm is able to cope with the different network conditions and constraints. Within the simulation environment, the differences in server temperatures and power consumption are limited. Further research is required to determine whether similar results are obtained while considering different system parameters.

Item Type:	Thesis (Bachelor's Thesis)
Degree programme:	Industrial Engineering and Management
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	15 Feb 2018 08:28
Last Modified:	15 Feb 2018 08:28
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/15270

Actions (login required)

View Item