Redis Service¶

Redis is an in-memory data structure store, used as a database, cache, and message broker.

Overview¶

Property	Value
Type	In-Memory Database
Default Port	6379
GPU Required	No
Container	`docker://redis:latest`

Quick Start¶

# Start Redis service
python main.py --recipe recipes/services/redis.yaml

# Run benchmark
python main.py --recipe recipes/clients/redis_benchmark.yaml --target-service redis_xxx

# Download results
python main.py --download-results

Recipe Configuration¶

Basic Recipe¶

# recipes/services/redis.yaml
service:
  name: redis
  description: "Redis in-memory database"

  container:
    docker_source: docker://redis:latest
    image_path: $HOME/containers/redis_latest.sif

  resources:
    nodes: 1
    ntasks: 1
    cpus_per_task: 4
    mem: "8G"
    time: "02:00:00"
    partition: cpu
    qos: default

  environment:
    REDIS_PORT: "6379"

  ports:
    - 6379

With Persistence¶

service:
  name: redis

  environment:
    REDIS_PORT: "6379"
    REDIS_APPENDONLY: "yes"
    REDIS_APPENDFSYNC: "everysec"

  # Bind data directory for persistence
  bind_mounts:
    - "$HOME/redis/data:/data"

With Monitoring¶

# recipes/services/redis_with_cadvisor.yaml
service:
  name: redis
  enable_cadvisor: true
  cadvisor_port: 8080
  # ... rest of config

Configuration Options¶

Option	Description	Default
`maxmemory`	Maximum memory limit	No limit
`maxmemory-policy`	Eviction policy	`noeviction`
`appendonly`	Enable AOF persistence	`no`
`save`	RDB snapshot intervals	Disabled

Benchmark Client¶

Single-Run Benchmark¶

# recipes/clients/redis_benchmark.yaml
client:
  name: redis_benchmark
  type: redis_benchmark

  parameters:
    clients: 50
    requests: 100000
    data_size: 256
    tests: "SET,GET,LPUSH,LPOP,SADD"
    output_file: "$HOME/results/redis_benchmark.json"

Parametric Benchmark¶

Run comprehensive sweeps across multiple parameters:

# recipes/clients/redis_parametric.yaml
client:
  name: redis_parametric
  type: redis_parametric_benchmark

  parameters:
    client_counts: [1, 10, 50, 100, 200]
    data_sizes: [64, 256, 1024, 4096]
    pipeline_depths: [1, 10, 50]
    tests: "SET,GET"

Run with:

./scripts/run_redis_parametric.sh

Benchmark Metrics¶

Metric	Description
`ops_per_second`	Operations per second
`latency_avg`	Average latency (ms)
`latency_p50`	Median latency
`latency_p95`	95th percentile
`latency_p99`	99th percentile

CLI Operations¶

Connect to Redis via redis-cli:

# Via SSH tunnel
ssh -L 6379:mel0182:6379 -N u103227@login.lxp.lu -p 8822

# Then locally
redis-cli
> SET key "value"
> GET key
> INFO

Performance Tuning¶

High Throughput¶

environment:
  REDIS_TCP_BACKLOG: "511"
  REDIS_MAXCLIENTS: "10000"

Memory Optimization¶

environment:
  REDIS_MAXMEMORY: "4gb"
  REDIS_MAXMEMORY_POLICY: "allkeys-lru"

Results Analysis¶

After running parametric benchmarks, analyze with:

# Generate plots
python analysis/plot_redis_results.py

# View summary
cat results/redis_parametric_summary.json | jq .

Generated plots include:

Throughput vs. clients
Latency distribution
Data size impact
Pipeline depth analysis