Web / mobile client
The browser, mobile app or partner system that originates requests.
Problem it solves: Every design starts with who calls the system, how often and with what latency expectation.
Where it sits: Start of every request path.
When to use
- Always: one per distinct caller population (web users, mobile, partner API).
When not to use
- Do not model internal services as clients; connect them as services.
Failure modes
- Client retries without back-off amplify outages (retry storms).
- Old app versions keep calling deprecated APIs.
Metrics to watch
- Offered load (req/s)
- Client-observed p95/p99 latency (ms)
- Error rate (%)
Alternatives
- Partner/API client
- Scheduled job as a traffic source