Recent versions of Ceph provide a feature known as "Persistent Write Log cache". From the Ceph documentation:

The Persistent Write Log Cache (PWL) provides a persistent, fault-tolerant write-back cache for librbd-based RBD clients.

This cache uses a log-ordered write-back design which maintains checkpoints internally so that writes that get flushed back to the cluster are always crash consistent. Even if the client cache is lost entirely, the disk image is still consistent but the data will appear to be stale.

This cache can be used with PMEM or SSD as a cache device. For PMEM, the cache mode is called replica write log (rwl). At present, only local cache is supported, and the replica function is under development. For SSD, the cache mode is called ssd.

We think this would be helpful in mitigating some of the performance issues seen when running an application that write small objects at a high transaction rate, such as etcd, on remote storage. We would like to investigate the effort involved in enabling this feature on master nodes.

Some questions:

• Can we enable this in a configuration file, or does it need to be part of the connection info (meaning changes to os-brick and potentially nova and/or cinder.
• When enabled, is there a significant performance improvement.
• ...