The Symbiosis of Storage and Workloads: Longhorn for Hyperconverged Bl... Jinhong Kim & Jangseon Ryu

This talk covers the transition of NAVER Corp from Ceph to Longhorn for hyperconverged block storage within their Kubernetes environment, which manages over one million Pods and more than 200,000 block persistent volumes. The speakers, Jinhong Kim and Jangseon Ryu, discuss the challenges associated with application-layer replication and fault isolation, which prompted the need for a hyperconverged approach. They detail how NAVER implemented coordinated scheduling to enhance the performance of storage and workloads while managing resource controls to mitigate the impact of storage operations on service I/O. The session highlights the effective use cases for different storage characteristics, showcasing NAVER's solutions to optimize performance and reliability.