RisingWave

View Current Viewing Revision #9 from 04/30/2023 10:48 p.m.

Streaming

RisingWave is an open-source distributed streaming database targeting real-time analytics and event-driven applications. It uses the incremental computation model to process streaming data with low latency. RisingWave implements a traditional change propagation framework to keep user-defined materialized views up-to-date. An incremental checkpoint mechanism is used to ensure data consistency. It also has an elastic multi-node architecture with separate data and compute nodes. RisingWave is built from scratch with Rust, and is wire compatible with PostgreSQL.

History

RisingWave Database is built by RisingWave Labs (formerly known as Singularity Data), a database systems startup founded in 2021 by former IBM researcher and Amazon Redshift engineer Yingjun Wu.

While working at Amazon Redshift, Wu noticed that existing database systems cannot process streaming data efficiently and existing streaming systems were too complicated for most companies to use. This observation motivated Wu to found RisingWave Labs with a mission to “democratize stream processing”.

Checkpoints

Non-Blocking Consistent

RisingWave uses the Chandy–Lamport algorithm to create consistent checkpoints.

To ensure that data is correct and consistent, read queries always fetch data from the most recent checkpoint. This means RisingWave does not ensure read-after-write consistency.

A local shared buffer is used to stage uncommitted write batches submitted by operators. The storage manager will notify all operators to commit their buffered writes into the shared storage when checkpoint trigger message has reached all operators.

Compression

Naïve (Page-Level) Prefix Compression

RisingWave applies both naive compression and prefix compression at block-level. It uses LZ4 and Zstd for naive compression.

Concurrency Control

Multi-version Concurrency Control (MVCC)

RisingWave employs Multi-version Concurrency Control (MVCC) in its cloud-native LSM-Tree-based storage engine. It creates one checkpoint at a time (as opposed to concurrent checkpointing), and performs read operation on the most-recently committed checkpoint. Read operation that immediately follows a write operation may not see the newest change if the two operations are performed on different compute nodes.

Data Model

Relational

RisingWave uses a relational data model. Relational tables are composed of a list of strongly-typed columns. All columns are implicitly nullable. RisingWave supports primitive data types of: boolean, integer, fixed-point and floating-point numbers, strings, and temporals. Composite data types of struct and list are also supported.