Olga Mirensky, Australia and New Zealand Banking Group, ANZx
Reducing cloud costs is one of the major concerns for tech companies today. One of the most cost effective ways to save on compute is to utilise Spot provisioning model. All major cloud vendors offer Spot Instances with up to 91% discount compared to on-demand prices and it’s tightly integrated in the respective vendor’s ecosystem, in particular in managed Kubernetes services like GKE, EKS and AKS. From our experience running a fleet of GKE clusters on Spot Instances, there’s much more to it than meets the eye. Losing capacity at a moment’s notice is only one part of the story and in this talk, we will delve into under-the-hood mechanisms of GKE Spot implementation, edge cases, and why teams collaboration and solid SRE principles are absolutely crucial in this environment.
Olga Mirensky, ANZx
![](https://www.usenix.org/sites/default/files/styles/speaker_photo/public/mirensky_olga_200x230.jpg?itok=lYvoleSs)
Olga is a Platform Engineer in Australia and New Zealand Banking Group focusing on building cloud infrastructure for the new digital bank. Her recent roles span years of experience working with Kubernetes of various shapes and flavours running on AWS, GCP and Azure, she has also developed managed OpenShift on Azure (ARO) while serving as a RedHat SRE. She loves exploring modern cloud native technologies and currently is experimenting with Cluster API, Cilium, eBPF and system performance.
![](https://www.usenix.org/sites/all/modules/usenix/usenix_files/images/usenix-locked.png)
author = {Olga Mirensky},
title = {Lessons Learned Running {GKE} Clusters on Spot Instances.},
year = {2023},
address = {Singapore},
publisher = {USENIX Association},
month = jun
}