KDS: A Kubernetes-based Dynamic Resource Scheduler for Real-Time Systems

Title
KDS: A Kubernetes-based Dynamic Resource Scheduler for Real-Time Systems

Speaker
Yang Li (Alex)

Abstract
Modularizing and containerizing for unified management and orchestration of services have increasingly become attractive models for industries. Developers can use Kubernetes to complete software deployment and DevOps efficiently. However, the load-balancing strategy provided by Kubernetes needs to be more effective in complex software deployment environments, especially in which services are real-time online.
In this topic, we will discuss three key questions first based on practical project experience: Why should we solve the load-balancing problem of real-time online services within the Kubernetes ecosystem? What are the difficulties in solving this problem? Furthermore, what solutions are there to solve this problem? Then we introduce KDS, a Kubernetes-based dynamic resource scheduler for real-time systems. The KDS has been deployed successfully in the production clusters and executes an adaptive scheduling strategy to make the cluster resources dynamic load-balancing. Through our long-term observation of KDS, KDS can run the scheduling strategy reliably and effectively according to the cluster state changes. In addition, our evaluation results showed that the performance overhead of KDS is acceptable.

Speaker's bio
My name is Yang Li, and My English name is Alex. I received Bachelor's and Master's degrees from Chongqing University of Posts and Telecommunications, Chongqing, China. Now I am a senior software development engineer at 58.COM, Beijing, China. My research interests and work experience include search engine techniques and highly available architecture design and development for Cloud-native applications.