Redis Performance Testing
Performance testing is an important way to evaluate Redis performance. This lesson covers using redis-benchmark for testing.
Introduction to redis-benchmark
redis-benchmark is Redis's built-in performance testing tool that can test various Redis operations.
Basic Usage
BASH
# Basic test (default 100000 requests, 50 concurrent clients)
redis-benchmark
# Specify host and port
redis-benchmark -h 127.0.0.1 -p 6379
# Specify request count and concurrency
redis-benchmark -n 100000 -c 100
Common Parameters
| Parameter | Description | Default |
|---|---|---|
-h |
Host address | 127.0.0.1 |
-p |
Port | 6379 |
-n |
Number of requests | 100000 |
-c |
Number of parallel clients | 50 |
-d |
Data size (bytes) | 3 |
-k |
Use keepalive | 1 |
-q |
Quiet output | false |
-t |
Commands to test | all |
-P |
Pipeline size | 1 |
Interpreting Test Results
Sample Output
TEXT
====== SET ======
100000 requests completed in 1.23 seconds
50 parallel clients
3 bytes payload
keep alive: 1
99.00% <= 1 milliseconds
99.80% <= 2 milliseconds
100.00% <= 3 milliseconds
81300.81 requests per second
====== GET ======
100000 requests completed in 1.15 seconds
50 parallel clients
3 bytes payload
keep alive: 1
99.00% <= 1 milliseconds
99.90% <= 2 milliseconds
100.00% <= 3 milliseconds
86956.52 requests per second
Key Metrics
| Metric | Description |
|---|---|
| requests per second | QPS (Queries Per Second) |
| 99.00% <= 1 milliseconds | 99% of requests completed within 1 ms |
| parallel clients | Number of concurrent clients |
| payload | Data size |
Common Test Scenarios
1. Testing SET Operations
BASH
# Test SET operation
redis-benchmark -t set -n 100000 -c 50
# Test SET with pipeline
redis-benchmark -t set -n 100000 -c 50 -P 10
2. Testing GET Operations
BASH
# Test GET operation
redis-benchmark -t get -n 100000 -c 50
3. Testing Specific Commands
BASH
# Test specific commands
redis-benchmark -t set,get,incr,lpush,rpush,lpop,rpop,sadd,hset -n 100000 -c 50
4. Testing Pipeline
BASH
# Use pipeline (batch send commands)
redis-benchmark -t set -n 100000 -c 50 -P 10
# Compare performance with and without pipeline
redis-benchmark -t set -n 100000 -c 50 -P 1
redis-benchmark -t set -n 100000 -c 50 -P 10
redis-benchmark -t set -n 100000 -c 50 -P 100
5. Testing Large Data
BASH
# Test with large data size
redis-benchmark -t set -n 100000 -c 50 -d 1024
# Test different data sizes
redis-benchmark -t set -n 100000 -c 50 -d 64
redis-benchmark -t set -n 100000 -c 50 -d 256
redis-benchmark -t set -n 100000 -c 50 -d 1024
Performance Optimization Tips
1. Use Pipeline
Pipelining can significantly improve performance by reducing network round trips:
BASH
# Without pipeline
redis-benchmark -t set -n 100000 -c 50 -P 1
# Result: ~80000 QPS
# With pipeline
redis-benchmark -t set -n 100000 -c 50 -P 10
# Result: ~200000 QPS
2. Use Batch Commands
Batch commands are more efficient than individual commands:
BASH
# Individual commands
redis-benchmark -t set -n 100000 -c 50
# Result: ~80000 QPS
# Batch commands (MSET)
redis-benchmark -t mset -n 100000 -c 50
# Result: ~150000 QPS
3. Tune Configuration
BASH
# Adjust max memory
redis-cli CONFIG SET maxmemory 256mb
# Adjust max client connections
redis-cli CONFIG SET maxclients 10000
# Disable persistence (for testing)
redis-cli CONFIG SET save ""
redis-cli CONFIG SET appendonly no
Performance Baseline Reference
Single Instance Performance
| Operation | QPS (no pipeline) | QPS (pipeline P=10) |
|---|---|---|
| SET | 80,000 | 200,000 |
| GET | 90,000 | 250,000 |
| INCR | 85,000 | 220,000 |
| LPUSH | 80,000 | 200,000 |
| LRANGE (100) | 30,000 | 80,000 |
💡 Note: Actual performance depends on hardware, network environment, data size, and other factors.
Custom Test Scripts
Using Lua Scripts
BASH
# Create a test script
cat > test.lua << 'EOF'
for i = 1, 100000 do
redis.call('SET', 'key:' .. i, 'value:' .. i)
end
EOF
# Execute the test
redis-cli --eval test.lua
Using Python
PYTHON
import redis
import time
r = redis.Redis(host='localhost', port=6379)
# Test SET operation
start = time.time()
for i in range(100000):
r.set(f'key:{i}', f'value:{i}')
end = time.time()
print(f'SET: {100000 / (end - start):.0f} QPS')
❓ FAQ
Q Do redis-benchmark results differ from production?
A Yes. redis-benchmark runs in an ideal environment. Actual production performance is affected by network, concurrency, data size, etc.
Q How do I test Redis cluster performance?
A Use redis-benchmark with specific node parameters, or use dedicated cluster testing tools.
Q What pipeline size is recommended?
A Typically 10-100, depending on network environment and data size.
📖 Summary
- redis-benchmark is Redis's built-in performance testing tool
- Key metrics: QPS, latency percentiles
- Pipelining can significantly improve performance
- Batch commands are more efficient than individual ones
- Actual performance depends on hardware, network, data size, etc.
📝 Exercises
- Basic test: Use redis-benchmark to test SET and GET operations
- Pipeline test: Compare performance with and without pipelining
- Optimization test: Try different configuration parameters and observe performance changes
Next Lesson
In the next lesson, we will learn Redis Pipelining, understanding the principles and usage of pipelining.