SSD cache (bcache)
This page originally covered SSD caching in general targeted for Fedora 20, including both dm-cache and bcache. In the processes it became clear that introducing basic bcache support was feasable for Fedora 20 and dm-cache had te be postponed until later.
Using recent kernel (3.9 and later) features for (fast) SSD caching of (slow) ordinary hard disks. Currently two kinds of SSD caching are available in the Linux kernel: bcache and dm-cache. In Fedora 20 basic bcache support is provided by the "bcache-tools" package, with small changes in util-linux, dracut and lvm2 as well. This page now provides suggestions on how to move forward in Fedora 22.
The Change described on this page is to implement bcache support in Anaconda, which allows easy installation of Fedora on "bcached" root-FS and support for existing (root or non-root) bcache partitions during Fedora installation.
Moving forward with dm-cache is not covered by this page, because the LVM team already plans to integrate dm-cache support in LVM2. See http://www.redhat.com/archives/dm-devel/2014-January/msg00072.html
- Name: Rolf Fokkens
- Email: email@example.com
- Name: Dave Lehman
- Email: dlehman AT fedoraproject
- Release notes owner:
Recent Linux kernels support the use of Solid State Drives as caches for rotational hard disks. Because the high cost per GB for SSD devices this feature may bring the best of both: fast end big yet affordable storage capacity. Linux kernel 3.9 introduced dm-cache, kernel 3.10 introduces bcache.
The introduction of bcache-tools in Fedora 20, and minor changes to util-linux, dracut and lvm2 allowed users to create SSD-cached block devices on which they could create bcache filesystems and/or LVM volumes. Creating a bcache root FS is possible using bcache-tools, but complicated because bcache requires a "fresh" block device destroying all data currently on it. Solving this involves adding bcache support in Anaconda. This covers both support for "manually created" (bcache-tools) bcached filesystems, and support for creating a bcached root FS. (Implementing bcache support in Anaconda can't be done by me, and needs to be done by the Anaconda team, hence I consider this a suggestion. Discussions on this topic with dlehman indicated an estimated effort of 1-2 weeks to implement).
Moving forward with dm-cache is not covered by this page because the LVM team already plans to integrate dm-cache support in LVM2. See http://www.redhat.com/archives/dm-devel/2014-January/msg00072.html. It should be noted that (in contrast to bcache) dm-cache support may not require Anaconda changes because it'll be an integral part of lvm2.
Benefit to Fedora
The availability of bcache-tools in Fedora 20 already brought the speed of SSD's to Fedora users, while also bringing the capacity of HDD's. This was limited however to expert Fedora users who feel comfortable using te command line.
The SSD caching support in Fedora 20 will become available to more Fedora users when Anaconda will support this in Fedora 22.
- Proposal owners: build and test blocks package (or include blocks in bcache-tools), support Anaconda team when needed, test Anaconda.
- Other developers: Build bcache support in Anaconda (by Anaconda team)
- Release engineering:
- Policies and guidelines:
All existing Fedora functionality remains as is.
How To Test
Testing involves doing a clean install of Fedora, while configuring in anaconda a cached root FS. After the installation the cache state should be checked (if it is operational or not.
- run benchmarks (see performance improvement over non cached storage)
- update the kernel (see if dracut builds a functional initramfs including caching
Users should experience improved storage performance after easy installation of Fedora in bcache devices.
Anaconda will depend on bcache tools
- Contingency mechanism: Revert bcache changes in Anaconda, revoke the blocks package.
- Contingency deadline: beta freeze
- Blocks release? No