Evaluating the energy consumption impact of a carbon aware autoscaling in microservice-based applications on the public cloud : a sustainability perspective
Gebreweld, Haben Birhane (2023)
Lataukset:
Diplomityö
Gebreweld, Haben Birhane
2023
School of Engineering Science, Tietotekniikka
Kaikki oikeudet pidätetään.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2023080894291
https://urn.fi/URN:NBN:fi-fe2023080894291
Tiivistelmä
Background: The growing awareness of the environmental impact of the ICT industry, including the software sector, and the emergence of sustainability policies and compliance requirements have fueled an increasing interest in sustainable application development and operation. Cloud computing has been widely acclaimed for its scalability, presenting prospects for sustainable software operation. Nonetheless, existing research is deficient in carbon-aware scaling strategies, especially in the context of microservice-based applications. Bridging this gap can lead to the establishment of environmentally conscious and efficient application scaling methods. This endeavor holds promise for advancing sustainability in the field of cloud computing.
Aim: The aim of this research is to empirically evaluate the energy consumption impact of Carbon-Aware Autoscaling on microservice-based applications in the Public Cloud.
Method: In this study, we conducted an empirical experiment to evaluate the impact of the autoscaling strategy on a microservice-based application. The experiment was designed to collect data on two dependent variables: energy consumption and response time. Through this approach, we aimed to gain insights into how the chosen autoscaling strategy affects these metrics.
Result: Our analysis resulted in three key discoveries regarding the impact of Carbon Aware Autoscaling on microservice-based applications. Primarily, it presents substantial energy cost efficiency, making it particularly advantageous in handling larger volumes of microservices. Secondly, Carbon Aware Autoscaling demonstrates notably lower energy consumption compared to HPA-based autoscaling, reducing energy usage by 42.9% on average on application level and 38.24% on average on microservice level, especially advantageous for microservices with heavy workloads. Thirdly, the autoscaler significantly impacts response time, showing an average increment of 30.68%, making it well-suited for low-priority workloads that do not require real-time processing. These findings provide essential insights for future research in the field of autoscaling strategies for microservice-based applications.
Conclusion: our analysis suggests the following insights: a) Extending the experiment to workload scheduling based on spatial and temporal carbon intensity values could yield valuable findings. b) Validating the effectiveness of the autoscaling approach with real-world microservices-based applications that do not require real-time processing is recommended. c) Mixing the two autoscaling strategies based on microservice workload may further improve energy consumption reduction and mitigate the impact on application response time. These recommendations offer potential avenues for future research and optimization in the field of autoscaling strategies for microservices-based applications.
Aim: The aim of this research is to empirically evaluate the energy consumption impact of Carbon-Aware Autoscaling on microservice-based applications in the Public Cloud.
Method: In this study, we conducted an empirical experiment to evaluate the impact of the autoscaling strategy on a microservice-based application. The experiment was designed to collect data on two dependent variables: energy consumption and response time. Through this approach, we aimed to gain insights into how the chosen autoscaling strategy affects these metrics.
Result: Our analysis resulted in three key discoveries regarding the impact of Carbon Aware Autoscaling on microservice-based applications. Primarily, it presents substantial energy cost efficiency, making it particularly advantageous in handling larger volumes of microservices. Secondly, Carbon Aware Autoscaling demonstrates notably lower energy consumption compared to HPA-based autoscaling, reducing energy usage by 42.9% on average on application level and 38.24% on average on microservice level, especially advantageous for microservices with heavy workloads. Thirdly, the autoscaler significantly impacts response time, showing an average increment of 30.68%, making it well-suited for low-priority workloads that do not require real-time processing. These findings provide essential insights for future research in the field of autoscaling strategies for microservice-based applications.
Conclusion: our analysis suggests the following insights: a) Extending the experiment to workload scheduling based on spatial and temporal carbon intensity values could yield valuable findings. b) Validating the effectiveness of the autoscaling approach with real-world microservices-based applications that do not require real-time processing is recommended. c) Mixing the two autoscaling strategies based on microservice workload may further improve energy consumption reduction and mitigate the impact on application response time. These recommendations offer potential avenues for future research and optimization in the field of autoscaling strategies for microservices-based applications.