Ceph continues to be the leading block-storage-of-choice for Openstack VM deployments. Latency-sensitive workloads such as databases remain an emerging Ceph use case - latency guarantees typically a challenge in shared VM storage environments as compute/networked-storage resources cease to scale as VM density increases. With growing flash footprint on clients, flash-based read/write caching has the potential to improve storage latency by reducing IO path dependency on the network.
We'll introduce an ordered, crash-consistent read/write-back cache native to the Ceph block layer. Blocks will be cached on client-side SSD/NVMe/Optane/NVRAM device. A shared cache mode will be supported which is an ideal fit for VM master image cache. A big feature-up from current object-based, DRAM-based cacher - flash increases cache size on packed hypervisor and improves cache warmup on restart. We'll delve into the high-level design and demonstrate performance gains with this new Ceph-RBD cache.