Declarative cluster configuration

Define distributed clusters directly in your Spice environment. Designate a master node and executor pool that register automatically with the scheduler—no external orchestration or complex setup required.

Parallel execution engine

Built on Apache Ballista and DataFusion, Spice distributes query stages across nodes for concurrent reads, joins, and aggregations. Each executor processes partitions independently and merges results in real time.

Automatic dependency management

Spice executors automatically detect and load dependencies—like embedding models or UDFs—from the master node. Runtime optimizers ensure each executor fetches only what’s required, reducing overhead.

Optimized query planning

A custom optimizer decomposes large scans into shuffle partitions and pushes projections closer to the data source. Less data movement, faster results, and lower network cost.