Private cloud / data center
Deploy the platform beside your catalogs, lakehouse services and supported compute runtimes.
Keep the platform and data inside your boundary.Architecture guide · Deployment
Your infrastructure,one lakehouse platform.
Deploy Realfuture in your environment and connect catalogs, storage, analytics and compute. Configure AI model access to fit your network and data policies.
Deployment models
Deploy within your enterprise boundary.
Choose a model based on data residency, runtime location, and operational ownership. Confirm final responsibilities during the architecture review.
Customer cloud / BYOC
Run Realfuture beside workloads in the cloud account your organization governs.
Use your cloud account and security controls.Hybrid deployment
Connect mixed runtimes while keeping compute placement and data residency under customer control.
Standardize across mixed infrastructure.Decision matrix
Decide control-plane placement, data residency, and runtime location first.
Reference patterns narrow the options; an architecture review validates networking, identity, high availability, and service ownership.
DimensionPrivate / data centerCustomer cloud / BYOCHybrid deployment
Data planeCustomer environmentCustomer cloud / BYOC accountCustomer environments
Runtime targetsYARN, Kubernetes, dedicatedKubernetes or dedicatedMultiple mixed targets
Data residencyCustomer boundaryCustomer cloud / BYOC boundaryPer connected environment
ConnectivityInternal platform-to-runtime accessAccess within the customer cloud accountExplicit network paths between environments
Best fitRegulated or on-premises estatesCloud-standardized teamsMulti-cloud and hybrid organizations
Architecture process
Resolve hard constraints before the POC.
A focused deployment review turns a generic demo into a meaningful test.
Map the runtime estate
Catalogs, compute, Paimon or Fluss storage, analytics engines, AI providers and network boundaries.
Assign ownership
Infrastructure, upgrades, support, security, and incident responsibilities.
Choose the topology
Platform, catalogs, storage, compute and AI provider connectivity, identity and availability.
Test a real workload
Evaluate delivery, observability, recovery, and day-two operations.
Compatibility and data
Check the components and the boundaries.
Managing an existing Flink runtime and using commercial runtime enhancements can have different requirements. Confirm the supported release before changing production jobs.
Review job dependencies, connectors and state compatibility. For each topology, identify where business records, previews, configuration, logs, metrics and credentials are processed, including any support or external AI service access.
Turn constraints into a deployable architecture.
Share your data stack, AI provider requirements and security boundaries to define a reference architecture and evaluation scope.