Kubernetes, the container orchestrator for cloud-deployed applications, offers automatic scaling for the application provider in order to meet the ever-changing intensity of processing demand. This auto-scaling feature can be customized with a parameter set, but those management parameters are static while incoming Web request dynamics often change, not to mention the fact that scaling decisions are inherently reactive, instead of being proactive. We set the ultimate goal of making cloud-based applications’ management easier and more effective. We propose a Kubernetes scaling engine that makes the auto-scaling decisions apt for handling the actual variability of incoming requests. In this engine various machine learning forecast methods compete with each other via a short-term evaluation loop in order to always give the lead to the method that suits best the actual request dynamics. We also introduce a compact management parameter for the cloud-tenant application provider to easily set their sweet spot in the resource over-provisioning vs. SLA violation trade-off. We motivate our scaling solution with analytical modeling and evaluation of the current Kubernetes behavior. The multi-forecast scaling engine and the proposed management parameter are evaluated both in simulations and with measurements on our collected Web traces to show the improved quality of fitting provisioned resources to service demand. We find that with just a few, but fundamentally different, and competing forecast methods, our auto-scaler engine, implemented in Kubernetes, results in significantly fewer lost requests with just slightly more provisioned resources compared to the default baseline.

中文翻译：

---

基于机器学习的Kubernetes Edge集群扩展管理

Kubernetes是用于云部署应用程序的容器协调器，它为应用程序提供商提供自动扩展功能，以满足不断变化的处理需求。可以使用参数集自定义此自动缩放功能，但是这些管理参数是静态的，而传入的Web请求动态通常会更改，更不用说缩放决策本质上是被动的，而不是主动的。我们设定了最终目标，使基于云的应用程序的管理更轻松，更有效。我们提出了一个Kubernetes伸缩引擎，该引擎使自动伸缩决策易于处理传入请求的实际可变性。在该引擎中，各种机器学习预测方法通过短期评估循环相互竞争，以便始终引导该方法最适合实际的请求动态。我们还为云租户应用程序提供商引入了一个紧凑的管理参数，可以轻松地在资源超额配置与SLA违规权衡之间设定自己的优势。我们通过对当前Kubernetes行为的分析建模和评估来激发我们的扩展解决方案。在模拟中以及对我们收集的Web跟踪进行测量时，都可以评估多预测缩放引擎和建议的管理参数，以显示将已调配的资源与服务需求相匹配的改进质量。我们发现，通过自动缩放引擎，只有几种但根本不同且相互竞争的预测方法，