Tuning

Configuration profiles optimized for different workloads.

Baseline Profiles

Default Profile (Balanced)

For general-purpose applications:

RedisModule.forRoot({
  clients: {
    host: process.env.REDIS_HOST,
    port: 6379,
  },
  plugins: [
    new CachePlugin({
      l1: { maxSize: 1000, ttl: 30000 },
      l2: { defaultTtl: 3600 },
    }),
    new LocksPlugin({
      defaultTtl: 30000,
      autoRenew: { enabled: true, interval: 10000 },
    }),
    new RateLimitPlugin({
      algorithm: 'sliding-window',
    }),
  ],
})

Parameter	Value	Rationale
L1 size	1000	Moderate memory, good hit rate
L1 TTL	30s	Balance freshness/efficiency
L2 TTL	1 hour	Standard cache lifetime
Lock TTL	30s	Typical operation duration
Lock renewal	10s	Renew at 1/3 TTL

Latency-Sensitive Profile

For APIs where response time is critical:

RedisModule.forRoot({
  clients: {
    host: process.env.REDIS_HOST,
    port: 6379,
    commandTimeout: 1000,  // 1s timeout
  },
  plugins: [
    new CachePlugin({
      l1: { maxSize: 5000, ttl: 60000 },  // Larger L1
      l2: { defaultTtl: 300 },             // Shorter L2
      stampede: { enabled: true },
    }),
    new LocksPlugin({
      defaultTtl: 10000,   // Shorter locks
      retry: { maxRetries: 1, initialDelay: 100 },  // Fast fail
      autoRenew: { enabled: false },
    }),
    new RateLimitPlugin({
      algorithm: 'token-bucket',  // Allows bursts
    }),
  ],
})

Parameter	Value	Rationale
Command timeout	1s	Fail fast
L1 size	5000	More in-memory hits
L1 TTL	60s	Longer local cache
Lock TTL	10s	Short operations only
Lock timeout	1s	Don't wait long

Throughput Profile

For batch processing and background jobs:

RedisModule.forRoot({
  clients: {
    host: process.env.REDIS_HOST,
    port: 6379,
    enableOfflineQueue: true,
  },
  plugins: [
    new CachePlugin({
      l1: { enabled: false },  // Skip L1 for throughput
      l2: { defaultTtl: 7200 },
    }),
    new LocksPlugin({
      defaultTtl: 300000,  // 5 min for long operations
      autoRenew: { enabled: true, interval: 60000 },
    }),
    new StreamsPlugin({
      consumer: {
        batchSize: 100,       // Large batches
        blockTimeout: 10000,  // Long poll
      },
    }),
  ],
})

Parameter	Value	Rationale
L1 cache	Disabled	Avoid memory pressure
Lock TTL	5 min	Long batch operations
Batch size	100	Reduce round trips
Block timeout	10s	Efficient long polling

Memory-Constrained Profile

For edge deployments or small containers:

RedisModule.forRoot({
  clients: {
    host: process.env.REDIS_HOST,
    port: 6379,
  },
  plugins: [
    new CachePlugin({
      l1: { maxSize: 100, ttl: 10000 },  // Tiny L1
      l2: { defaultTtl: 300 },            // Short TTL
    }),
    new LocksPlugin({
      autoRenew: { enabled: false },  // Save resources
    }),
  ],
})

Parameter	Value	Rationale
L1 size	100	Minimal memory
L1 TTL	10s	Quick expiration
L2 TTL	5 min	Reduce Redis memory
Auto-renew	Disabled	Fewer background tasks

Parameter Reference

Cache Parameters

Parameter	Range	Impact
l1.maxSize	100-10000	Memory usage, hit rate
l1.ttl	5s-60s	Freshness vs hits
l2.defaultTtl	60s-86400s	Redis memory, staleness
stampede.enabled	true/false	Protection vs complexity

Lock Parameters

Parameter	Range	Impact
defaultTtl	5s-300s	Safety vs flexibility
timeout	0-30s	Wait time vs responsiveness
autoRenew.interval	TTL/3-TTL/2	Reliability vs overhead

Stream Parameters

Parameter	Range	Impact
batchSize	1-1000	Throughput vs latency
blockTimeout	1s-30s	Responsiveness vs efficiency
maxRetries	1-10	Reliability vs DLQ growth

Tuning Process

1. Start with Default Profile
2. Monitor Key Metrics
   • Cache hit rate
   • Lock wait time
   • Operation latency
3. Identify Bottlenecks
   • Low hit rate → Increase TTL or L1 size
   • High lock contention → Reduce scope or TTL
   • High latency → Enable L1, reduce timeouts
4. Adjust One Parameter at a Time
5. Validate with Load Testing

Anti-Patterns

Anti-Pattern	Problem	Solution
L1 TTL > L2 TTL	Serving stale data	L1 TTL < L2 TTL
Lock TTL < operation time	Lost locks	Measure operation, add margin
No command timeout	Hung requests	Set reasonable timeout
Huge batch sizes	Memory spikes	Balance batch size

Next Steps

• Monitoring — Track tuning effectiveness
• Connection Management — Connection-level tuning