Expand description
Distributed query execution — plan queries across the swarm.
The distributed planner assigns cells to nodes based on cache locality and available capacity. The coordinator writes a query manifest to storage, executes local tasks, polls for partial results from workers, merges them, and returns the final result.
Structs§
- Cell
Info - Information about a cell needed for query planning.
- Node
Info - Information about a node from the world view.
- Planned
Task - A task assigned to a specific node in a distributed query.
- Query
Manifest - Query manifest written to storage for distributed execution.
Enums§
Functions§
- cleanup_
query - Clean up query directory after completion.
- create_
manifest - Create a new query manifest.
- extract_
alive_ nodes - Extract node information from the world view (helper for runtime). This will be called by the runtime layer that has access to WorldView.
- generate_
sql_ fragment - Generate SQL fragment for a query (simplified for v1).
- manifest_
path - Query manifest path in storage.
- partial_
result_ path - Partial result path for a node.
- plan_
query - Plan a query: decide whether to execute locally or distribute across nodes.
- read_
manifest - Read query manifest from storage.
- read_
partial_ result - Read partial result from Arrow IPC file.
- write_
manifest - Write query manifest to storage.
- write_
partial_ result - Write partial result as Arrow IPC file.