Skip to content

Latest commit

 

History

History
119 lines (85 loc) · 3.58 KB

README.md

File metadata and controls

119 lines (85 loc) · 3.58 KB
Raw
Starscourge Radahn

Radahn

Radahn is a simple yet fast in-memory key-value store that speaks Redis serialization protocol.

Overview · Installation · Usage · Benchmarks

Overview

Radahn is designed to be simple. It supports a tiny fraction of Redis commands, only the most useful ones, so that there are minimum overhead costs. Further, it fully utlizes modern multi-core chip hardware by using the Shared Nothing Architecture. As a result, Radahn can provide more than 2x performance gains, in both throughput and latency, over Redis (see benchmarks).

Supported commands

Use case

Radahn is suitable to be used as a simple in-memory cache server that only needs to operate GET, SET and DEL commands. However, if your use case allows for multiple commands to be executed once, consider Redis as it supports pipelining.

Installation

Radahn can be quickly installed using Docker.

Prerequisites

  • docker

Steps

  • Build Docker image
docker build -t radahn:latest .
  • Run Docker container
docker run --rm -p 6369:6369 --name radahn -d radahn:latest

Radahn can also be installed by building from source. Currently, only the Linux platform is supported due to the use of its epoll API.

Prerequisites

  • gcc
  • make

Steps

  • Run tests (optional)
make check
  • Build the binary
make
  • Run the Radahn server
./server

Usage

Once the Radahn server is running, you can use any Redis client to connect and send RESP-compatible commands. Remember to the change the TCP port that your client connects to. Radahn by default listens on port 6369.

The following demo uses redis-cli

redis-cli -p 6369
127.0.0.1:6369> PING
PONG
127.0.0.1:6369> SET hello world EX 20
OK
127.0.0.1:6369> GET hello
"world"
127.0.0.1:6369> DEL hello
(integer) 1
127.0.0.1:6369> GET hello
(nil)
127.0.0.1:6369> QUIT

Benchmarks

Used the benchmarking utility that comes alongwith a default Redis installation called redis-benchmark. Ran the benchmark for commands GET and SET using 10 million requests and 10,000 random keys on both Radahn and Redis servers.

redis-benchmark -n 10000000 -q -t get,set --threads 16 -r 10000

Here are the results:

Radahn

SET: 139822.98 requests per second, p50=0.351 msec
GET: 156269.53 requests per second, p50=0.311 msec

Redis

SET: 67649.84 requests per second, p50=0.703 msec
GET: 73795.30 requests per second, p50=0.655 msec

Although Radahn is faster in basic benchmark, Redis can be configured to use pipelining to execute multiple commands at once. Radahn does not support pipelining currently. Here's an excerpt from Redis' website:

By default, redis-benchmark does not represent the maximum throughput a Redis instance can sustain. Actually, by using pipelining and a fast client (hiredis), it is fairly easy to write a program generating more throughput than redis-benchmark.