From Fedora Project Wiki

LVM Cache Logical Volumes

Summary

LVM can now use fast block devices (e.g. SSDs and PCIe Flash) to improve the performance of larger but slower block devices. These hierarchical or layered logical volumes are called "Cache Logical Volumes" in LVM.

Owner

Current status

Detailed Description

LVM is now capable of using fast block devices (e.g. SSDs) as write-back or write-though caches for larger slower block devices. Users can create cache logical volumes to improve the performance of their existing logical volumes or create new cache logical volumes composed of a small and fast device coupled with a large and slow device. These cache logical volumes can be used with most LVM segment types, including RAID 1/4/5/6/10, linear, stripe and thin pools.

Benefit to Fedora

Users will have the advantage of fast block device speeds with the capacity of larger but slower block devices.

Scope

  • Proposal owners:

The LVM team must deliver the lvm2 package implementing cache LV (already included in release 2.02.106)

  • Other developers: N/A (not a System Wide Change)

The Anaconda team must develop a UI for configuring cache LVs during installation. If Anaconda support is not provided, users will have to configure cache LVs after installation or by dropping into a command line. Also, Anaconda could fail if installing a new OS onto an existing cache LV if support is not provided.

The dracut team must provide boot support. If dracut does not provide support, cache LVs will not be usable as root devices.

  • Release engineering: N/A (not a System Wide Change)
  • Policies and guidelines: N/A (not a System Wide Change)

Upgrade/compatibility impact

  • Nothing changes for users following update path.
  • The logical volumes using cache will become unavailable from systems using older kernels and LVM. The cache will have to be dropped to grant access for those legacy systems.

How To Test

It is not strictly necessary to have SSDs or other fast block devices. You can use any block device as the cache for any other block device - even if they are the same device, as long as there is enough space! While you may not see the benefits of this arrangement without a fast device, it is still possible to check for correct functionality (i.e. can the cache LV be created, removed, resized, renamed, or changed? is there any corruption? etc).

If Anaconda has support for LVM cache logical volumes, users will be able to configure them at install time. Otherwise, it is possible to convert existing logical volumes to cache logical volumes once the system has been installed.

Information on how to setup cache logical volumes can be found in the lvm(8) man page.

User Experience

Users will enjoy the increased speed of disk storage while keeping capacity of HDD.

Dependencies

  • Anaconda: Implementation of UI
  • Dracut: Module to Detect and install required configuration files and kernel modules

Contingency Plan

  • LVM2: Changes are already present upstream. If severe bugs are found, cache can be disabled here.
  • Dracut: There is already a hardcoded dm-cache installation. If the detection is not done in time, it will work as is.
  • Anaconda: Cache will work even without UI except it will not be configurable by regular users without using command line. Users can also setup cache LVs after installation.
  • LVM2: We will need an ability to drop the cache should the cache device fail or become corrupted. (This can be done using Live CD.)
  • Contingency deadline: beta freeze
  • Blocks release? No
  • Blocks product? -

Documentation

  • Command line: Cache section of the upstream lvm(8) man page.

Release Notes