A singleton database process centralizes operations on a shared resource, serializing access through one server. This prevents races and limits unbounded concurrent I/O, but the server can become a bottleneck. section 7.3.4
Choose among these designs according to the resource and workload:
- Bypass the process when no long-lived state, reusable resource, or synchronization requirement exists; a plain module can improve parallelism. section 7.3.4
- Keep the central process but spawn one-off workers when requests share state yet can execute independently. section 7.3.4
- Keep handlers short and fast so serialized coordination does not prevent many workers from running concurrently. section 7.3.4
Visualizes requests passing through the central server while worker processes perform database operations concurrently. section 7.3.4
