Workflow Mode vs Server Group Mode
BizFirstGO execution nodes have two fundamental deployment modes: Workflow Mode (nodes execute inside a Flow Studio workflow) and Server Group Mode (nodes run as persistent services consumed by workflows and agents over HTTP). This guide explains both, when to use each, and how Octopus agents fit into the server group picture.
Two Deployment Modes
Workflow Mode
An execution node is instantiated and executed as a step inside a Flow Studio workflow. The workflow engine controls the node's lifecycle — start, execute, pass data, terminate.
Best for: sequential business processes, approval chains, data transformation pipelines.
Server Group Mode
An execution node runs as a long-lived HTTP service registered in a Server Group. Callers — workflows, agents, and other services — invoke it over HTTP without knowing the implementation details.
Best for: shared infrastructure, AI inference endpoints, high-throughput stateless services.
Conceptual Comparison
| Aspect | Workflow Mode | Server Group Mode |
|---|---|---|
| Lifecycle | Created per workflow execution, destroyed on completion | Long-lived process, shared across many callers |
| State | Stateful — carries workflow context and variables | Stateless by convention; shared state via external store |
| Invocation | Workflow engine calls ExecuteAsync(nodeContext) | HTTP calls to REST endpoints |
| Scalability | Scales with workflow parallelism | Independently scalable (Kubernetes HPA) |
| Discovery | Flow Studio node palette, drag-and-drop | Server Group registry, URL-based |
| Reuse | Per-workflow — one instance per run | Shared — many workflows call the same server |
What This Guide Covers
| Page | Topic |
|---|---|
| Workflow Mode | Node executor lifecycle, BaseNodeExecutor pattern, when to use workflow nodes |
| Server Group Mode | Server Group registration, HTTP invocation, load balancing, health probes |
| Server Node as a Service | Why server nodes are powerful shared services; patterns for high-throughput scenarios |
| AI Inference Server Node | Serving AI models as server nodes — GPU allocation, batching, token streaming |
| Decision Framework | 10-question guide to choosing workflow vs server group mode |
| Satellite Nodes | Edge-deployed execution nodes that report back to the central server group |
| Octopus as a Server Node | Deploying the Octopus agent runtime itself as a server node inside a Server Group |