一句话：提出 Snape 框架，用约束强化学习动态混合 Spot 与按需虚拟机，在保持高可用性的同时降低云成本。
云提供商常以低价提供未充分利用的资源（如 Spot 虚拟机），但其可用性较低，客户难以在成本与资源可用性之间做出权衡。本文提出 Snape（Spot On-demand Perfect Mixture），一个通过动态混合按需虚拟机与 Spot 虚拟机来优化成本与资源可用性的智能框架。基于真实生产跟踪数据的详细刻画，作者验证了 Spot 虚拟机的驱逐在一定程度上是可预测的。Snape 进一步利用约束强化学习在线调整混合策略。在不同配置下的实验表明，与仅使用按需虚拟机相比，Snape 实现了 44% 的成本节省，同时保持 99.96% 的可用性，比仅使用 Spot 虚拟机高 2.77%。
Cloud providers often have resources that are not being fully utilized, and they may offer them at a lower cost to make up for the reduced availability of these resources. However, customers may be hesitant to use such offerings (such as spot VMs) as making trade-offs between cost and resource availability is not always straightforward. In this work, we propose Snape (Spot On-demand Perfect Mixture), an intelligent framework to optimize the cost and resource availability by dynamically mixing on-demand VMs with spot VMs. Through a detailed characterization based on real production traces, we verify that the eviction of spot VMs is predictable to some extent. Snape also leverages constrained reinforcement learning to adjust the mixture policy online. Experiments across different configurations show that Snape achieves 44% savings compared to using only on-demand VMs while maintaining 99.96% availability, which is 2.77% higher than using only spot VMs.
@inproceedings{paperbot10093,
title = {Snape: Reliable and Low-Cost Computing with Mixture of Spot and On-Demand VMs},
author = {Fangkai Yang and Lu Wang and Zhenyu Xu and Jue Zhang and Liqun Li and Bo Qiao and Camille Couturier and Chetan Bansal and Soumya Ram and Si Qin and Zhen Ma and Íñigo Goiri and Eli Cortez and Terry Yang and Victor Rühle and Saravan Rajmohan and Qingwei Lin and Dongmei Zhang},
booktitle = {ASPLOS},
year = {2023},
doi = {10.1145/3582016.3582028}
}