Single service comparison

Apigee separates a hosted control plane (proxy definition, policies, analytics, developer portal, monetization) from a runtime that can be deployed as Google-hosted or Apigee hybrid in a customer GKE cluster; the data plane scales and patches independently with runtime versions. Huawei APIG instead delivers a single managed gateway-as-a-service where the instance, gateway specifications, and environments are provisioned together, and it additionally bundles microservice-gateway capabilities (service discovery, gray release, gRPC) that Apigee leaves to adjacent tooling. API import is supported both sides, but Apigee's OpenAPI/SmartAPI import plus proxy-revision versioning model has no identical counterpart in APIG.

Policy and developer-experience boundaries differ materially. Apigee provides an extensible policy model (quota, spike arrest, OAuth, JWT, threat protection, JavaScript/Java callouts, custom shared flows) and a built-in developer portal with app registration and monetization/billing flows for API products. APIG addresses the same concerns through request throttling by app/API/user/IP, custom and third-party authorizers, and WAF interconnection for web-attack protection, but it offers richer microservice traffic governance and lighter built-in monetization tooling. Treat APIG as the gateway equivalent and use ROMA Connect or CodeArts where Apigee's integration or developer-portal features are load-bearing.

Scaling, observability, and operational responsibility differ. Apigee's runtime and metrics are operated by Google with TPS governed by tier and autoscaling, and analytics are a first-class managed product. APIG publishes single-gateway TPS up to 80,000 with selectable gateway specifications, integrates Cloud Eye/CTS for monitoring and audit, and expects the customer to choose instance size and region placement; HA is regional. Cross-region and on-prem backend exposure work through APIG dedicated gateways and VPC peering, whereas Apigee handles multi-region routing via its own hosted or hybrid runtime topology rather than per-region gateway instances.

Assess each Apigee API group against APIG capabilities before committing to migration. Inventory proxies, shared flows, custom policies (JavaScript/Java), OAuth/JWT providers, developer portal apps, monetization plans, and target backends, then map them to APIG concepts: API groups, environments, custom authorizers, throttling policies, and WAF or IAM for security. Flag Apigee hybrid or Apigee Edge features, monetization, and DeepTrace analytics as gaps that need compensating services (ROMA Connect, CodeArts) or external tooling; do not assume feature-by-feature parity for the control plane.

Migrate API definitions and configuration by importing OpenAPI/OAS specs into APIG groups and rebuilding backend routing, request/response transformation, and throttling policies per API. Re-implement custom Apigee policies as APIG custom authorizers on FunctionGraph, plug-in policies, or backend-side logic, validating signature, OAuth, and JWT flows against equivalent identity providers. Confirm VPC and backend reachability using dedicated gateways or cross-VPC backend access, and recreate environment promotion (test/prod) through APIG environments and release management rather than Apigee proxy revisions.

Validate functional, non-functional, and security behavior before cutover. Replay representative traffic or run shadow/staged traffic against the APIG gateway, comparing latency, error codes, throttling behavior, and OAuth/JWT outcomes with Apigee baselines. Configure Cloud Eye alarms and CTS audit trails to mirror Apigee analytics dashboards, and run developer-portal and app-key provisioning flows through the APIG console or ROMA Connect to confirm subscription and credentialing parity.

Address gaps and the cost-model change explicitly. Huawei typically bills APIG by gateway instance specification and request volume plus WAF/FunctionGraph usage, whereas Apigee bills by product tier plus included API calls and proxy runtime; recompute TCO for peak TPS, request volume, retention, and cross-region traffic before production. Explicitly cover developer portal, monetization, custom policy portability, and analytics depth as residual gaps, and plan a phased cutover with DNS/traffic-weighting to allow rollback, because no one-click Apigee-to-APIG migration path exists.