A stylized illustration of tiny computer components.
Enlarge / OpenZFS 2.0.0 brings a ton of latest options and efficiency enhancements to each Linux and BSD platforms.

Aurich Lawson

This Monday, ZFS on Linux lead developer Brian Behlendorf revealed the OpenZFS 2.0.0 launch to GitHub. Together with numerous new options, the announcement brings an finish to the previous distinction between “ZFS on Linux” and ZFS elsewhere (for instance, on FreeBSD). This transfer has been a very long time coming—the FreeBSD group laid out their aspect of the roadmap two years in the past—however that is the discharge that makes it official.


The brand new OpenZFS 2.0.0 launch is already out there on FreeBSD, the place it may be put in from ports (overriding the bottom system ZFS) on FreeBSD 12 techniques, and would be the base FreeBSD model within the upcoming FreeBSD 13. On Linux, the state of affairs is a little more unsure and relies upon largely on the Linux distro in play.

Customers of Linux distributions which use DKMS-built OpenZFS kernel modules will are inclined to get the brand new launch slightly shortly. Customers of the better-supported however slower-moving Ubuntu most likely will not see OpenZFS 2.0.0 till Ubuntu 21.10, practically a yr from now. For Ubuntu customers who’re keen to dwell on the sting, the favored however third-party and individually maintained jonathonf PPA would possibly make it out there significantly sooner.

OpenZFS 2.0.0 modules could be constructed from supply for Linux kernels from 3.10-5.9—however most customers ought to stick with getting prebuilt modules from distributions or well-established builders. “Far past the overwhelmed path” is just not a phrase one ought to typically apply to the file system that holds one’s treasured information!

New options

  • Sequential resilver—rebuilding degraded arrays in ZFS has traditionally been very totally different from typical RAID. On practically empty arrays, the ZFS rebuild—generally known as “resilvering”—was a lot sooner, as a result of ZFS solely wants to the touch the used portion of the disk, slightly than cloning every sector throughout your entire drive. However this course of concerned an abundance of random I/O—so on extra practically full arrays, typical RAID’s extra pedestrian block-by-block whole-disk rebuild went a lot sooner. With sequential resilvering, ZFS will get the very best of each worlds: largely sequential entry, whereas nonetheless skipping unused parts of the disk(s) concerned.
  • Persistent L2ARC—one in every of ZFS’ most compelling options is its superior learn cache, generally known as the ARC. Methods with very massive, very popular working units may also implement an SSD-based learn cache known as L2ARC, which populates itself from blocks within the ARC nearing eviction. Traditionally, one of many largest points with L2ARC is that though the underlying SSD is persistent, the L2ARC itself is just not—it turns into empty on every reboot (or export and import of the pool). This new characteristic permits information within the L2ARC to stay out there and viable between pool import/export cycles (together with system reboots), vastly rising the potential worth of the L2ARC system.
  • Zstd compression algorithm—OpenZFS presents clear inline compression, controllable at per-dataset granularity. Historically, the algorithm mostly used has been lz4, a streaming algorithm providing comparatively poor compress ratio however very mild CPU loading. OpenZFS 2.0.0 brings assist for zstd—an algorithm designed by Yann Collet (the creator of lz4) which goals to supply compression much like gzip, with CPU load much like lz4.

    These graphs are a bit troublesome to observe—however primarily, they present zstd reaching its targets. Throughout compression (disk writes), zstd-2 is extra environment friendly than even gzip-9, whereas sustaining excessive throughput.

    In comparison with lz4, zstd-2 achieves 50 p.c larger compression in return for a 30 p.c throughput penalty. On the decompression (disk learn) aspect, the throughput penalty is barely larger, at round 36 p.c.

    Take into accout, the throughput “penalties” described assume negligible bottlenecking on the storage medium itself. In follow, most CPUs can run rings round most storage media (even comparatively gradual CPUs and quick SSDs). ZFS customers are broadly accustomed to seeing lz4 compression speed up workloads in the actual world, not gradual them down!

  • Redacted replication—this one’s a little bit of a brain-breaker. For example there are parts of your information that you just do not need to again up utilizing ZFS replication. First, you clone the dataset. Subsequent, you delete the delicate information from the clone. Then, you create a bookmark on the dad or mum dataset, which marks the blocks which modified from the dad or mum to the clone. Lastly, you’ll be able to ship the dad or mum dataset to its backup goal, together with the --redact redaction_bookmark argument—and this replicates the nondelicate blocks solely to the backup goal.

Extra enhancements and modifications

Along with the key options outlined above, OpenZFS 2.0.0 brings fallocate assist, improved and reorganized man pages, larger efficiency for zfs destroy, zfs ship, and zfs obtain, extra environment friendly reminiscence administration, and optimized encryption efficiency. In the meantime, some occasionally used options—deduplicated ship streams, dedupditto blocks, and the zfs_vdev_scheduler module possibility—have all been deprecated.

For a full listing of modifications, please see the unique launch announcement on GitHub at https://github.com/openzfs/zfs/releases/tag/zfs-2.0.0.


Please enter your comment!
Please enter your name here