CURVE: a High-performance Software-Defined storage for Cloud Native applications Xiang Dong Architect, Netease Hangzhou Research InstituteAgenda � Why develop storage Design objectives Achievements Easy operation and maintenance ○ High performance ● CEPH ○ Complex and Large amount of codes ○ system maintenance are difficult in corner cases ○ Performance requirements cannot be metAgenda Why develop Storage for KVM / Kubernetes / iSCSI Design objectives ● Thin provisioning storage pools ● High performance ● Easy maintenance ● Cloud NativeAgenda Why develop storage Design objectives � Achievements

Curve High performance Cloud native Distributed storage system https://www.opencurve.io/Agenda • What is Curve • Use Cases • CurveBS • Key Features • Comparing to Ceph • CurveFS • Comparing Block Storage (CurveBS) • CurveBS: a high performance cloud native distributed block storage • Curve File System (CurveFS) • CurveFS: a high performance cloud native file systemUse Cases • Container on-prem OSSCurveBS • high performance • mainly used for (SSD, three replicas) • csi / storage class for kubernete, nbd for HOST/VMPerformance (vs. Ceph RBD)Performance (vs. Ceph RBD)CurveBS Features

storage system • High performance • Easiser operation • Cloud native • Curve is composed of two parts • Curve Block Storage (CurveBS) • CurveBS: a high performance cloud native distributed distributed block storage • Curve File System (CurveFS) • CurveFS: a high performance cloud native file system (Based on CurveBS / S3 compatible Storage)Operator capability Level OPERATOR CAPABILITY LEVEL CURVE Curve failure reports(disk / pods / mds/ chunkserver) to be viewed in dashboard • Gather Curve performance /delay reports to be viewed in dashboard

High performance Cloud native Distributed storage system https://www.opencurve.io/Agenda • CurveBS Architecture • CurveBS Topology • CurveBS Data Organization • MetaData Server (MDS) • ChunkServer ChunkServer • Client • CurveBS IO processing flow • CurveBS Performance considerations • Cloud NativeAgenda • CurveFS Architecture • CurveFS Data Organization • CurveFS file Organization • write • Reduce write magnification for file new write • Cooperate with Alibaba to support high-performance polardb for postgresql using CurveBS • Cloud-native support for CurveBS • CurveBS clusters

possible to allocate space frequently at the beginning, but after the allocation is complete, performance recovers. 3. The Client queries the ChunkServer for the leader ChunkServer node of the copyset used to store file data. Curve-fuse Supports Object storage by S3 apis and CurveBS CurveBS performance considerations 1. CurveBS uses raft as a consistency protocol which requires majority replicas