One of the design goals around Leveled was to make the production of anti-entropy trees easier to achieve. Originally I intended this as a backend feature (i.e. it would be only available in Riak + leveled).
Over the past few months I've worked on KV Index Tictac Tree - which is an AAE library, that takes some of the ideas from the Leveled/AAE work, and makes them generally available across other backends. The library is now ready to share as a working prototype: https://github.com/martinsumner/kv_index_tictactree The intention is to use this as a building block to provide a drop-in replacement to existing Riak AAE for Riak 3.0. Anticipated features are: - Will run in parallel to existing AAE, so can support a controlled migration; - Can rapidly provide Merkle Trees merged over coverage plans - to make inter-cluster and inter-database replication much faster and easier to implement; - Can be run without a dedicated AAE keystore if using the leveled backend, but still provide equivalent features when using other backends; - Supports coverage folds across the AAE store per-bucket for reporting queries; - Will be "always on", so no loss of anti-entropy capability during AAE rebuilds; - Provides access for Repair functions to the clock differences between objects, so intelligent repair decisions can be made (e.g. no longer replicate out of date objects due to hash differences). Over the next month I want to get this plugged into Riak, and run some performance tests. Any feedback appreciated. Regards Martin P.S. I will be talking about this at CodeBEAM Stockholm - http://www.erlang-factory.com/codebeamsto2018
_______________________________________________ riak-users mailing list riak-users@lists.basho.com http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
