Ray

A minimalistic key-value storage.

Disclaimer: this project is made solely for educational purposes and for the fun of the author. PRs and issues are welcome, but don't expect them to be addressed in a timely manner.

About Ray

Ray is persistent key-value storage with linearizability guarantee written in async Rust. It works over gRPC, has a simple proto api, extensive debug logging, metrics collection and even a Graphana dashboard to visualize them.

The project consists of:

• Server binary - rayd;
• Rust client library suitable for use with Tokio runtime;
• Command-line client - ray;
• Benchmarking tool - ray-benchmark.

Running rayd

$ cargo run --release --bin rayd

This will run rayd in default configuration. It will:

• Create journal directory for write-ahead log;
• Create snapshots directory for storage snapshots;
• Create rayd.debug.log file (if does not exist) and write debug logs to it;
• Listen on port 39172 for client connections.

Options are supplied to rayd via config file. Example config can be found in examples/config.yml. To run rayd with config file, do

$ cargo run --release --bin rayd -- -c example/config.yml

Using ray

ray is a command-line tool that allows you to interact with rayd manually. For example:

$ cargo run --bin ray -- set my_key my_value
$ cargo run --bin ray -- get my_key

This will try to connect to rayd assuming it is listening on localhost:39172. Use --address and --port keys to connect to a different address and port.

Running ray-benchmark

Benchmarking tool supports two modes: read and write. Example usage:

$ cargo run --release --bin ray-benchmark -- --tasks 1000 --delay 10000 --key-len 8 read

This will run 1000 tasks, each making a read request of a random 8-byte key approximately every 10000 mcs.