Exploring the Longhorn Data Plane: An Overview and Enhancements

By Jing-Wei Su

議題

Exploring the Longhorn Data Plane: An Overview and Enhancements

TR 211 [[ new Date( '2023-07-30 03:30:00+00:00' ).toLocaleDateString('ja', {year: 'numeric', month: '2-digit', day: '2-digit'}) ]] [[ new Date( '2023-07-30 03:30:00+00:00' ).toLocaleTimeString('zh-Hant', {hour12: false, hour: '2-digit', minute:'2-digit'}) ]] ~ [[ new Date( '2023-07-30 04:00:00+00:00' ).toLocaleTimeString('zh-Hant', {hour12: false, hour: '2-digit', minute:'2-digit'}) ]] zh-tw
加入行事曆 加入關注 加入關注 已關注

Longhorn is a distributed block storage system that has been purpose-built for Kubernetes, making it an ideal cloud-native solution. As an incubating project of the Cloud Native Computing Foundation (CNCF), its main goal is to provide persistent and highly available storage for stateful workloads. In this presentation, we will be exploring Longhorn's architecture and specifically, the data path. Despite offering a wealth of features, such as snapshots, backups, and recurring jobs, there is still room for improvement in the data plane's performance. To address this, a data engine is integrated with Storage Performance Development Kit (SPDK) in the upcoming release, which improves the I/O data path, resulting in better latency and throughput. Additionally, we will introduce the Userspace Block Device Driver (ublk) as a frontend to enhance performance and reduce latency. This feature is particularly useful for edge devices that use legacy data engines.

講者

Jing-Wei Su

Jing-Wei Su

Jing-Wei Su is a Staff Software Engineer at SUSE from the end of 2021. He is interested in storage technology. He is currently working on the CNCF Longhorn project.

Kubernetes Community Day Taiwan KSJLLL general (30mins)