Difference between revisions of "CephCephFS"
Jump to navigation
Jump to search
(Created page with "== Introduction == Three ways to access storage: Object, Block, File Object RGW Block RBD File: CephFS Object storage: scles well. simple consistancy model, one object at ...") |
(No difference)
|
Latest revision as of 10:34, 22 October 2014
Introduction
Three ways to access storage: Object, Block, File
Object RGW Block RBD File: CephFS
Object storage: scles well. simple consistancy model, one object at a time, clients are simple, to replationships between objects, one name per object, scale out achieved through mapping objects to nodes
POSIX filesystems are hard to scale out, due to lcocking, multiple clients accessing the same file, clients are stateful, loss of data and damage whole subtree, inodes depend on each other,
CephFS - Dynamically balcance scale out meta data, POSIX compatability, subtree snapshots, recursive stats CephFS components: client, server (MDS), storage (OSD)