Real-time, zero-regression performance metrics comparing azums against industry alternatives.
Measures atomic single-job insertion speed across memory, Redis, SQLite, and PostgreSQL storage engines.
Elapsed milliseconds between job enqueue and idle worker wake-up via LISTEN/NOTIFY and PubSub.
Throughput scaling under lock contention when multiple worker tasks concurrently lease job batches.
High-speed event publishing, consumer group offset reading, and sequence ACK throughput.
Head-to-head comparison of throughput, CPU usage, and latency across queue architectures.
| Framework / Queue | Backend Engine | Enqueue Throughput | Idle CPU Usage | Wake-up Latency |
|---|---|---|---|---|
| azums (Memory) | In-Memory Broadcast | 380,000 ops/sec | 0.0% | < 0.1 ms |
| azums (Redis) | Redis 7.0 Multiplex | 145,000 ops/sec | 0.0% | < 0.8 ms |
| azums (Postgres) | LISTEN / NOTIFY | 42,000 ops/sec | 0.0% | 1.2 ms |
| Traditional Polling Queue | PostgreSQL Polling | 8,500 ops/sec | 12.4% (wasted) | 500 ms (default poll) |
| Python Celery | RabbitMQ / Redis | 3,200 ops/sec | 2.1% | 15.0 ms |