1. Scenario Overview
In a connectivity management platform, the business system should call one capability. Underneath, that call still has to reach China Mobile, China Unicom, and China Telecom, and those APIs do not look alike. The operator API gateway sits between the SimtoLink IoT Cloud Platform and each operator. It owns routing, protocol adaptation, rate limits, circuit breaking, and observability, so many networks and many protocols become one call chain the platform can reuse.
The usual pattern is to connect the gateway to the operator CMP or to an industry gateway. The channel is chosen from the ICCID or the number range. SOAP, REST, and proprietary payloads are translated. Timeouts, rate limits, and operator maintenance windows get a fallback and a retry, so one bad endpoint does not stop batch activation or suspend and resume. Operator system integration is about what you connect. This page is about how a call gets in, how it gets out, and how it stays up.
2. Core capabilities
Routing and load across operators
Route by ICCID range, project, customer pool, or region. When more than one operator can take the call, balance by weight or round robin. A primary channel can fail over to a standby, or the call can wait in a queue, so the business keeps moving when one operator API is down.
One authentication model
Certificates, keys, IP allowlists, and signature algorithms for each operator live on the gateway. The business side uses the platform credential only. Rate limits and concurrency caps stop a mistake, or a bad client, from hammering an operator API, which is what operator security reviews and audits expect.
Health checks and the circuit breaker
Probe each operator API on a schedule. Repeated failures open the circuit: calls fail fast and an alert goes out. When the endpoint recovers, a half-open probe checks it before traffic returns, which stops a cascade. Failed work from async jobs and batch commands lands in a retry queue and runs again on a backoff.
Monitoring and diagnosis
Each call keeps a request ID, the duration, the operator return code, and a redacted payload summary. Those records show up next to multi-operator management. Success rate and P99 latency can be sliced by API, operator, or customer, which is the evidence for capacity planning and an SLA.
3. The gateway and operator integration
Operator system integration is the joint test and the credentials for each operator API. The gateway is how those connections are served, with high availability, at runtime. After both are in place, you do not keep a connection pool, a retry policy, and an alert rule per operator. The platform operates them together.
4. Where teams use it
- Routing a mixed fleet: Millions of SIMs span China Mobile, China Unicom, and China Telecom. The gateway routes on the ICCID range. When an operator API enters a maintenance window, traffic moves to the standby channel and the business call does not change.
- Batch suspend and resume: A large daily batch goes out to several operators at once. If an API degrades, the circuit breaker parks the work in a queue. After the window, retries follow a backoff, so failed jobs do not pile up as permanent errors.
- A baseline for each operator API: P99 latency and success rate are tracked continuously. When one operator slows down, the gateway alerts and shifts routing weight. Those figures are what you take into an SLA discussion with the operator.
5. How to put it in place
When a new operator or a new API is added, start with a canary route on the gateway: a small share of traffic first, then the rest. Rotate keys and certificates, and keep a change log. For critical commands such as batch activation or suspend and resume, set both a synchronous timeout and an asynchronous callback, so interactive calls and large batches each have a path. With open API integration, the shape is operator gateway, platform core, then the enterprise API. Each layer has a job, and the boundary is explicit.
6. What the gateway changes
The operator API gateway takes the multi-network wiring out of every product team, makes each call easier to trust and easier to trace, and isolates a bad operator API so the rest can recover. At fleet scale, that layer is part of the platform, not an optional add-on.