Persistent write log cache

Edit online

IBM Storage Ceph cluster, Persistent Write Log (PWL) cache provides a persistent, fault-tolerant write-back cache for librbd-based RBD clients.

PWL 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. If the client cache is lost entirely, the disk image is still consistent but the data appears stale. You can use PWL cache with persistent memory (PMEM) or solid-state disks (SSD) as cache devices.

For PMEM, the cache mode is replica write log (RWL) and for SSD, the cache mode is (SSD). Currently, PWL cache supports RWL and SSD modes and is disabled by default.

Primary benefits of PWL cache are:

  • PWL cache can provide high performance when the cache is not full. The larger the cache, the longer the duration of high performance.

  • PWL cache provides persistence and is not much slower than RBD cache. RBD cache is faster but volatile and cannot guarantee data order and persistence.

  • In a steady state, where the cache is full, performance is affected by the number of I/Os in flight. For example, PWL can provide higher performance at low io_depth, but at high io_depth, such as when the number of I/Os is greater than 32, the performance is often worse than that in cases without cache.

Use cases for PMEM caching are:

  • Different from RBD cache, PWL cache has non-volatile characteristics and is used in scenarios where you do not want data loss and need performance.

  • RWL mode provides low latency. It has a stable low latency for burst I/Os and it is suitable for those scenarios with high requirements for stable low latency.

  • RWL mode also has high continuous and stable performance improvement in scenarios with low I/O depth or not too much inflight I/O.

Use case for SSD caching is:

  • The advantages of SSD mode are similar to RWL mode. SSD hardware is relatively cheap and popular, but its performance is slightly lower than PMEM.