With recent development efforts to add various "data compliance" mechanisms (such as Data Tiering, BitRot detection, dedup, etc.) to GlusterFS, a common scalable, pluggable & crash consistent filesystem notification infrastructure forms the basis for such techniques. In this session, Venky Shankar explains how Gluster uses it's journaling infrastructure to provide "pure" callback based notification facility via a set of easy to use APIs available as a part of shared library.
Log based synchronous & asynchronous (lazy) replication techniques in GlusterFS have already benefited in many aspects by relying on filesystem journals for data synchronization. The core infrastructure is extended by adding support for data classification (hot/cold data), multi-consumer model, improved performance, out-of-band notifications, etc., yet keeping the design simple and elegant.
This presentation would help users/developers to understand the low level details of the notification infrastructure of GlusterFS and help them understand, use and contribute to it's upstream development.
