Rasputin DB :globewithmeridians:

a distributed transactional store

triumvirs: stability, performance, and composability

key types

1. table: transactional namespaced KV backed by RocksDB
2. log: namespaced sequential messages with configurable retention policy
3. object: utilizes system VFS, useful for CDN type workloads

each type supports (possibly-disjoint) transactional range-based operations

client semantics

• read/write/delete a range
• subscribe: receive a linearized mutation stream on a specified range of a collection
• watch: receive a notification (at most once semantics, it may never come even if a mutation has occurred) when a mutation occurs on a range of a collection

replication modes (per-collection)

1. consensus: for use where loss of acked writes is unacceptable, and throughput is willing to be sacrificed
2. async: for high-throughput operations which can tolerate a finite window of data loss in the event of a master failure

roadmap

• [x] mio event loops
• [x] leadership algorithm for a range
• [x] rocksdb persistence layer
• [x] consensus algorithm based on parts of Raft and Paxos
• [ ] meta table that contains information about collections and replicas
• [ ] mutable membership over consensus algorithm
• [ ] lexicographic resharding of keyspace