Abstract
Modern cloud applications are expected to continuously provide adequate performance, withstanding changing workloads, heterogeneous hardware, and unpredictable infrastructure failures. Autoscaling can automatically provision resources to match performance goals but may suffer from slower reaction times and risks of over-provisioning. Brownout mechanisms, on the other hand, empower applications with the ability to quickly dim out optional features, freeing computational resources to serve core functionalities with the desired performance level. However, modifying an application to include brownout capabilities may require invasive changes to the codebase and the need to expose ad-hoc interfaces to coordinate the interaction of the brownout dimmers and autoscaling actions, avoiding interferences that may destabilize the system. In this paper, we report on our preliminary results on the design of an application-agnostic control theoretical solution to integrate scaling and dimming capabilities at the orchestrator level. We implemented a prototype of our controller on top of Kubernetes and HAProxy to empower generic applications with coordinated autoscaling and brownout capabilities by dynamically controlling the number of active replicas and per-user access to optional API endpoints.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Qu, C., Calheiros, R.N., Buyya, R.: Auto-scaling web applications in clouds: a taxonomy and survey. In: ACM Comput. Surv. 51(4) (2018). https://doi.org/10.1145/3148149
Chen, T., Bahsoon, R., Yao, X.: A survey and taxonomy of self-aware and self-adaptive cloud autoscaling systems. ACM Comput. Surv. 51(3) (2018). https://doi.org/10.1145/3190507
Klein, C., Maggio, M., Årzén, K.-E., Hernández-Rodriguez, F.: Brownout: building More robust cloud applications. In: ICSE, pp. 700–711. ACM Press, Hyderabad (2014).https://doi.org/10.1145/2568225.2568227
Xu, M., Dastjerdi, A.V., Buyya, R.: Energy efficient scheduling of cloud application components with brownout. IEEE Trans. Sustain. Comput. 1(2), pp. 40–53 (2016). arXiv: 1608.02707. https://doi.org/10.1109/TSUSC.2017.2661339
Wang, C., et al.: Effiective capacity modulation as an explicit control knob for public cloud profitability. ACM Trans. Auton. Adapt. Syst. 13(1), 21–225 (2018). https://doi.org/10.1145/3139291
Åström, K.J., Hägglund, T., Astrom, K.J.: Advanced PID control, vol. 461. ISA (2006)
Filieri, A., et al.: Control strategies for self-adaptive software systems. ACM Trans. Auton. Adapt. Syst. 11(4), 241–2431 (2017). https://doi.org/10.1145/3024188
Maggio, M., Papadopoulos, A.V., Filieri, A., Hoffmann, H.: Automated control of multiple software goals using multiple actuators. In: ESEC- FSE. ESEC/FSE 2017. ACM, August 2017. https://doi.org/10.1145/3106237.3106247
Maggio, M., Klein, C., Årzén, K.-E.: Control strategies for predictable brownouts in cloud computing. In: IFAC Proceedings Volumes. 19th IFACWorld Congress, vol. 47, no. 3, pp. 689–694, January 2014. https://doi.org/10.3182/20140824-6-ZA-1003.00669
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2020 Springer Nature Switzerland AG
About this paper
Cite this paper
Kotegov, I., Filieri, A. (2020). Towards Coordinated Autoscaling and Application Brownout at the Orchestrator Level. In: Muccini, H., et al. Software Architecture. ECSA 2020. Communications in Computer and Information Science, vol 1269. Springer, Cham. https://doi.org/10.1007/978-3-030-59155-7_21
Download citation
DOI: https://doi.org/10.1007/978-3-030-59155-7_21
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-59154-0
Online ISBN: 978-3-030-59155-7
eBook Packages: Computer ScienceComputer Science (R0)