resp-benchmark 0.1.4

No project description provided

resp-benchmark

resp-benchmark is a benchmark tool for testing databases that support the RESP protocol, such as Redis, Valkey, and Tair. It offers both a command-line interface and a Python library.

Installation

Requires Python 3.8 or higher.

pip install resp-benchmark

Usage

Command-Line Tool

resp-benchmark --help

Python Library

from resp_benchmark import Benchmark

bm = Benchmark(host="127.0.0.1", port=6379)
bm.flushall()
bm.load_data(command="SET {key sequence 10000000} {value 64}", count=1000000, connections=128)
result = bm.bench("GET {key uniform 10000000}", seconds=3, connections=16)
print(result.qps, result.avg_latency_ms, result.p99_latency_ms)

Custom Commands

resp-benchmark supports custom test commands using placeholder syntax like SET {key uniform 10000000} {value 64} which means the SET command will have a key uniformly distributed in the range 0-10000000 and a value of 64 bytes.

Supported placeholders include:

• {key uniform N}: Generates a random number between 0 and N-1. For example, {key uniform 100} might generate key_0000000099.
• {key sequence N}: Sequentially generates from 0 to N-1, ensuring coverage during data loading. For example, {key sequence 100} generates key_0000000000, key_0000000001, etc.
• {key zipfian N}: Generates according to a Zipfian distribution (exponent 1.03), simulating real-world key distribution.
• {value N}: Generates a random string of length N bytes. For example, {value 8} might generate 92xsqdNg.
• {rand N}: Generates a random number between 0 and N-1. For example, {rand 100} might generate 99.
• {range N W}: Generates a pair of random numbers within the range 0 to N-1, with a difference of W, used for testing *range* commands. For example, {range 100 10} might generate 89 99.

Best Practices

Benchmarking zset

# 1. Load data
resp-benchmark --load -n 1000000 -P 10 "ZADD {key sequence 1000} {rand 1000} {value 8}"
# 2. Benchmark
resp-benchmark "ZRANGEBYSCORE {key uniform 1000} {range 1000 10}"

Benchmarking Lua Scripts

redis-cli 'SCRIPT LOAD "return redis.call('\''SET'\'', KEYS[1], ARGV[1])"'
resp-benchmark "EVALSHA d8f2fad9f8e86a53d2a6ebd960b33c4972cacc37 1 {key uniform 100000} {value 64}"

Differences with redis-benchmark

When testing Redis with resp-benchmark and redis-benchmark, you might get different results due to:

1. redis-benchmark always uses the same value when testing the set command, which does not trigger DB persistence and replication. In contrast, resp-benchmark uses {value 64} to generate different data for each command.
2. redis-benchmark always uses the same primary key when testing list/set/zset/hash commands, while resp-benchmark generates different keys using placeholders like {key uniform 10000000}.
3. In cluster mode, redis-benchmark sends requests to each node, but all requests target the same slot on every node.