This is general knowledge or an external-engine reference, not an Apollo architecture or implementation claim. See Apollo's architecture status for target terminology and verified implementation boundaries.
Q53: How do you design scene management?
Core Conclusion
Scene management is not "loading a map"; it is "managing the complete lifecycle of a scene from creation, to hosting entities, to synchronization, to switching, to recycling".
What really needs to be designed clearly first is:
- The boundary between scenes and instances
- How entities within a scene are organized
- How scene switching and cross-scene migration work
- When scene resources and logical state are recycled
If these boundaries are unclear, the scene system gets tangled up with instances, AOI, synchronization, and resource management.
One: First Distinguish Several Concepts
1. Scene Template
Describes static content, for example:
- Map resources
- Spawn points
- Monster spawn points
- Navigation data
2. Scene Instance
Describes runtime state, for example:
- Current players and monsters
- Mechanism state
- Temporary events
3. Space Partitioning or Zoning
This is a hosting and scaling layer concern; it is not entirely the same as the scene itself.
Separating these three layers makes later extension much clearer.
Two: What Scene Management Is Usually Responsible For
At least it includes:
- Creating and destroying scene instances
- Managing the lifecycle of entities within the scene
- Maintaining scene-level rules and state
- Coordinating scene switching and migration
For instances and battlefields, these responsibilities are even heavier.
Three: Scene Switching Is a High-Risk Path
When a player moves from one scene into another, it is usually not just changing a coordinate.
You also need to handle:
- Leaving the old scene
- Admission to the new scene
- Entity migration or rebuilding
- AOI reset
- Client state synchronization
If the path has no explicit phases, scene switching easily produces black screens, duplicate entities, or lost state.
Four: The Scene Instance Lifecycle Must Be Explicit
It can commonly be divided into:
- Creation
- Activation
- Running
- Idle
- Recycling
Different scene types have different strategies:
- Main cities usually stay resident
- Instance scenes are usually created and recycled on demand
Five: Entity Organization Determines Management Cost
A scene usually contains:
- Players
- Monsters
- Drops
- Mechanisms
- Temporary effect objects
The scene management layer has to answer:
- Who is responsible for creating them
- Who is responsible for destroying them
- Who maintains visibility and synchronization boundaries
If every subsystem manages them on its own, scene consistency quickly gets out of control.
Six: The Scene System and the Resource System Should Also Be Decoupled
Server-side "scene management" cares more about logical state, while client-side "scene loading" cares more about resource presentation.
The two are related, but they cannot be merged into one.
The server side focuses on:
- State and entities
- Scene rules
- Synchronization boundaries
The client side focuses on:
- Map resources
- Models and effects
- Render loading
Seven: A More Solid Design in Engineering
A common approach is:
- Use scene templates to define static data
- Use scene instances to maintain runtime state
- Scene switching goes through an explicit state machine
- Empty scenes are recycled by policy
That way main cities, instances, battlefields, and other types can all land in one unified framework.
Eight: Common Misconceptions
1. Scene Management Is Map Loading
No. The server side cares more about scene instances and entity lifecycles.
2. Scene Switching Is Teleporting
No. It involves sessions, entities, synchronization, and client state rebuilding.
3. Destroying a Scene Immediately Once It Has No Players Is Always Best
Not necessarily. You also have to weigh reconnection, recycling cost, and whether delayed entry is needed.
References
- Various MMO / ARPG scene instance and map-switch management practice materials
