-
Notifications
You must be signed in to change notification settings - Fork 0
[Archive] A Carbon‐aware Workload Dispatcher in Cloud Computing System
JaeKyeong-Kim edited this page Sep 23, 2024
·
1 revision
Editor: JaekyeongKim
- conference: 2023 IEEE 16th International Conference on Cloud Computing (CLOUD)
- paper title: A Carbon-aware Workload Dispatcher in Cloud Computing System
- authors: Bahreini, Tayebeh and Tantawi, Asser and Youssef, Alaa
- job-level과 load-level 모두에서 Job placement and scheduling을 통해 클라우드 데이터 센터의 탄소 발자국을 최소화하기 위한 근사 알고리즘을 소개함.
- 근사 알고리즘은 마감 기한, 요구되는 컴퓨팅 자원, 실행 시간을 포함한 세 가지 특징을 기반으로 QoS와 계산 자원을 위반하지 않으며 job들의 실행 시기를 최적화하여 탄소 감축을 달성함.
- 지리적으로 분산된 여러 데이터센터에서 동시에 진행하는 job 단위 탄소 인지 클라우드 컴퓨팅은 이 연구를 다른 워크로드 스케줄링 연구와 차별화함.
- LP-relaxation 기반의 근사 알고리즘을 통해 확장성을 유지하며 roll-dice 기법에 기반한 Sample-based Random Round 알고리즘을 사용해 시간 효율적으로 NP-hard 문제의 정답 값에 가까운 정확도의 알고리즘을 제공함.
- 여러 데이터센터에서 진행하는 동시간적 job 스케줄링 접근 방식은 시공간적 워크로드 스케줄링을 진행하는 점에서 기존 연구들과 차별화함.
- 세 종류의 job 특성에 중점을 두고 있지만, 다른 특성은 연구되지 않았음. 연구에서 사용된 데이터셋에는 다양한 특성이 존재하므로, 이 네 가지 특성이 다른 특성보다 우선시된 이유에 대한 설명이 필요함.
- 논문에서 제시하는 SRR-APX 알고리즘은 auto-scaling 작업이 클라우드에서 적용되지 않을 때에 효과적임. 하지만, 현재 대부분의 클라우드 서비스에서는 오토스케일링 기능이 지원됨.
- 계산 자원에 대한 위반이 감지되면, fail되는 job들이 발생함. 계산 자원에 대한 유연성이 떨어지는 시스템에서 유용한 알고리즘이나, 해당 환경에서 QoS 문제가 발생함.
T. Bahreini, A. Tantawi and A. Youssef, "A Carbon-aware Workload Dispatcher in Cloud Computing Systems," 2023 IEEE 16th International Conference on Cloud Computing (CLOUD), Chicago, IL, USA, 2023, pp. 212-218, doi: 10.1109/CLOUD60044.2023.00032.
This Wiki is maintained by the Data Science Lab at Dong-A University. Our mission is to provide comprehensive resources and research archives on Sustainable Cloud Computing.
If you have any questions or need further information, please feel free to contact us
© 2024 Data Science Lab, Dong-A University. All rights reserved.