vocabulary: name: Trino Vocabulary description: >- Domain vocabulary for Trino, the open-source distributed SQL query engine. Covers core concepts, data types, connectors, query execution, and cluster management. version: "480" created: "2026-05-03" modified: "2026-05-03" terms: - term: Catalog definition: >- A named data source registered in Trino's connector framework. Each catalog maps to a single connector instance and provides access to schemas and tables within that data source. Example: hive, iceberg, postgresql, mysql. tags: - Architecture - Data Sources - term: Connector definition: >- A Trino plugin that provides integration with a specific data storage or processing system. Connectors implement the Trino SPI and expose tables, schemas, and SQL operations over an underlying data source such as Hive, PostgreSQL, Kafka, or Iceberg. tags: - Architecture - Integrations - term: Coordinator definition: >- The single Trino node that accepts client SQL queries, parses and plans them, and distributes execution stages to worker nodes. The coordinator hosts the /v1/statement REST endpoint and the Trino Web UI. tags: - Architecture - Cluster Management - term: Worker definition: >- A Trino node that executes query fragments (splits) assigned by the coordinator. Workers process data locally and exchange intermediate results with other workers via the exchange protocol. tags: - Architecture - Cluster Management - term: Split definition: >- The unit of parallelism in Trino. Each split represents a subset of data to be read from a connector (e.g., an HDFS file block, a Kafka partition segment). Workers execute splits in parallel to produce query results. tags: - Execution - Parallelism - term: Stage definition: >- A sequential phase of query execution that can run in parallel across multiple workers. The query plan is divided into stages separated by data exchange operations. tags: - Execution - Query Plan - term: Task definition: >- An instance of a stage running on a specific worker node. Tasks process splits and produce operator results that are fed to the next stage. tags: - Execution - term: Operator definition: >- A single logical operation within a task, such as table scan, filter, project, aggregation, join, or exchange. Operators form a pipeline within a task. tags: - Execution - Query Plan - term: QueryResults definition: >- The JSON response object returned by POST /v1/statement and GET nextUri. Contains the query ID, column metadata, data rows, execution statistics, and optionally a nextUri for polling more results. tags: - API - Data Model - term: nextUri definition: >- A URL in the QueryResults response that clients must GET to retrieve the next batch of results. Clients follow nextUri until it is absent, signaling query completion. DELETE nextUri cancels the query. tags: - API - Protocol - term: Session definition: >- A logical client context for query execution. Sessions hold the current catalog, schema, time zone, locale, resource group tags, and session properties set via X-Trino-Session headers or USE statements. tags: - Authentication - Configuration - term: Session Property definition: >- A named configuration parameter that modifies query execution behavior within a session. Set via X-Trino-Session header (key=value) or SET SESSION statement. Examples: query_max_memory, optimize_hash_generation, join_distribution_type. tags: - Configuration - term: Resource Group definition: >- A named administrative unit used to manage concurrency and memory limits for groups of queries. Resource groups are matched using client tags (X-Trino-Client-Tags) and enforce queue limits and priorities. tags: - Administration - Resource Management - term: Fault-Tolerant Execution definition: >- A Trino execution mode that checkpoints intermediate results to durable storage, enabling query recovery from node failures without restarting from scratch. Useful for very long-running analytics queries. tags: - Reliability - Execution - term: Pushdown definition: >- An optimization where Trino translates SQL operations (filters, projections, aggregations, joins) into connector-specific operations executed in the data source itself, reducing data transfer and improving performance. tags: - Optimization - Performance - term: Dynamic Filtering definition: >- A runtime optimization that uses join build-side data to prune splits and rows on the probe side, dramatically reducing I/O for selective join queries. tags: - Optimization - Performance - term: Exchange definition: >- The mechanism by which intermediate query results are transferred between stages and workers. Exchanges can be local (within a worker) or remote (across workers). tags: - Execution - Architecture - term: Type definition: >- Trino's data type system. Includes primitive types (BOOLEAN, TINYINT, SMALLINT, INTEGER, BIGINT, REAL, DOUBLE, DECIMAL, VARCHAR, CHAR, VARBINARY, DATE, TIME, TIMESTAMP, INTERVAL) and complex types (ARRAY, MAP, ROW, JSON, UUID, IPADDRESS). tags: - Data Types - Schema - term: Iceberg definition: >- An open table format for large analytic datasets. Trino's Iceberg connector supports ACID transactions, schema evolution, partition evolution, and time travel queries on data stored in object storage (S3, GCS, ADLS). tags: - Connectors - Data Lakes - term: Hive Connector definition: >- Trino's connector for data stored in Hive-compatible formats (ORC, Parquet, Avro, RCFile, JSON) on HDFS, S3, or other storage. The most widely used Trino connector for data lake workloads. tags: - Connectors - Data Lakes