Welcome to Hibari
Hibari is a distributed, ordered key-value store with strong consistency guarantee. Hibari is written in Erlang and designed for being:
- Fast, Read Optimized: Hibari serves read and write requests in short and predictable latency. Hibari has excellent performance especially for read and large value operations
- High Bandwidth: Batch and lock-less operations help to achieve high throughput while ensuring data consistency and durability
- Big Data: Can store Peta Bytes of data by automatically distributing data across servers. The largest production Hibari cluster spans across 100 of servers
- Reliable: High fault tolerance by replicating data between servers. Data is repaired automatically after a server failure
Hibari is able to deliver scalable high performance that is competitive with leading open source NOSQL (Not Only SQL) storage systems, while also providing the data durability and strong consistency that many systems lack. Hibari’s performance relative to other NOSQL systems is particularly strong for reads and for large value (> 200KB) operations.
As one example of real-world performance, in a multi-million user webmail deployment equipped with traditional HDDs (non SSDs), Hibari is processing about 2,200 transactions per second, with read latencies averaging between 1 and 20 milliseconds and write latencies averaging between 20 and 80 milliseconds.
Unlike many other distributed databases, Hibari uses “chain replication methodology” and delivers distinct features.
- Ordered Key-Values: Data is distributed across “chains” by key prefixes, then keys within a chain are sorted by lexicographic order
- Always Guarantees Strong Consistency: This simplifies creation of robust client applications
- Compare and Swap (CAS): key timestamping mechanism that facilitates “test-and-set” type operations
- Micro-Transaction: multi-key atomic transactions, within range limits
- Custom Metadata: per-key custom metadata
- TTL (Time To Live): per-key expiration times