Master Thesis MSTR-2022-09

BibliographyPiontek, Tobias: CO2 aware job scheduling for data centers.
University of Stuttgart, Faculty of Computer Science, Electrical Engineering, and Information Technology, Master Thesis No. 9 (2022).
74 pages, english.
Abstract

Data centers consume large amounts of power around the world. It is estimated that over 1% of global energy consumption is used for powering data centers. Therefore data centers have some potential to reduce global CO2 output. This thesis therefore introduces a novel practical scheduler implementation for saving CO2 emissions on a cluster by shifting load in time. Therefore a custom CO2 power grid efficiency scheduler is developed. The implementation is written for kubernetes, as it is widely used open source cloud orchestration tool. Different architectural solutions to implement a scheduler inside kubernetes are discussed, to find a good approach for realization for this specific cause. The scheduler predicts future CO2 emissions by using historical data and shifts job in time to CO2 efficient power grid times. For comparison the implemented scheduler is tested against the default kubernetes scheduling implementation with multiple different scenarios that were built by using real world workload log data. The implementation presented achieved an average CO2 emission reduction between 0.5% and 2.0%. The scheduler CO2 reduction is similar to Googles Borg scheduler implementation.

Full text and
other links
Volltext
Department(s)University of Stuttgart, Institute of Architecture of Application Systems
Superviser(s)Aiello, Prof. Marco; Haghshenas, Dr. Kawsar
Entry dateMay 31, 2022
   Publ. Institute   Publ. Computer Science