Optimizing Longhorn for high performance hardware

Longhorn is an open-source, Cloud-native volume manager, which implements its own distributed block storage system. It is a complete and independent software defined storage solution, handling internally all aspects related to capacity management, performance, fault-tolerance, as well as interfacing with both Kubernetes and the end user. Longhorn is an actively-developed and mature software, however our installations have revealed that it currently lacks the ability to take advantage of the hardware performance available in local servers that feature high-speed solid state disks and high-bandwidth network connectivity. This presentation investigates a series of performance optimizations that have been engineered in the Longhorn's core that collectively allow the system to achieve an order of magnitude better IOPS and bandwidth.