Design and Implementation of the FreeBSD Operating System, 2nd Edition

Chapter 10. The Zettabyte Filesystem

10.1 Introduction

The Zettabyte filesystem is generally referred to as simply ZFS [Bonwick et al., 2003]. It is in a class of filesystems that never overwrite existing data. A benefit of never overwriting is that snapshots (read-only) and clones (writable) are easy and cheap. Many of them can be created with no performance hit.

ZFS has the property that the on-disk filesystem state is never inconsistent. Filesystem changes are accumulated in memory. Periodically, all the changes are gathered up and written to disk. When all the changes are on stable storage, ZFS makes a checkpoint of the new filesystem state. The checkpoint is made by doing a single write to update the uberblock to reference the new filesystem ...