Data Migration & Validation glossary¶
Terms used across AIM DMV documentation, listed alphabetically.
Accepted transformation¶
A declarative rule in a validation workflow that whitelists a specific source-to-target value pair on a column (or columns matching a regex pattern). AIM DMV does not report matching diffs as mismatches. Requires L3 row validation. See Accepted transformations.
Anti-locking / query modifiers¶
Optional SQL hints appended to source queries to reduce locking on busy source tables. Set queryModifiers with objectModifier (appended after the table in FROM) and selectModifier (inserted after SELECT). User-configured hints are off by default. See Anti-locking and query modifiers.
AIM DMV¶
Snowflake AIM Data Migration and Validation: the combined framework for moving data into Snowflake and validating that it matches the source. Uses an Orchestrator, Workers, and the SNOWCONVERT_AI metadata database.
BCP¶
Bulk Copy Program extraction strategy for SQL Server. Optional high-throughput bulk export when use_bcp is enabled and the bcp utility is installed on the Worker host.
Cell validation¶
The L3 step that compares individual column values when row fingerprinting detects a mismatch. Results are stored in CELL_VALIDATION_RESULTS.
Custom metrics¶
Workflow-wide overrides for L2 aggregate metrics per data type. Set validationCustomMetrics at the workflow root with source and target arrays. See Customizing normalization and metrics.
Compute pool¶
Snowflake object that hosts Snowpark Container Services. You create a compute pool before deploying Orchestrator or Worker services. See Deploying on Snowpark Container Services.
DUPLICATE_
L3 row validation results for duplicated index-column keys: DUPLICATE_SOURCE (the key appears more than once on the source), DUPLICATE_TARGET (more than once on the target), and DUPLICATE_BOTH_SIDES (duplicated on both sides). See Validation levels and result codes.
External access integration¶
Snowflake object that grants an SPCS service permission to reach external hosts defined in a network rule. Required for Workers that connect to source databases or download drivers at runtime.
Extraction strategy¶
How a Worker reads data from the source during migration. Platform-specific options include regular (default ODBC/JDBC query), UNLOAD (Redshift), WRITE_NOS and TPT (Teradata), and BCP (SQL Server).
Hybrid table¶
Snowflake table type used for SNOWCONVERT_AI metadata when your account supports them (for example the task queue). When Hybrid Tables are not available, AIM DMV creates equivalent metadata as standard tables, which works but is slower. See Prerequisites on the overview page.
L1 / L2 / L3¶
Validation levels: L1 schema, L2 metrics (aggregates), L3 row and cell comparison. See Data validation.
MISMATCH¶
L3 row validation result when source and target values differ after comparison. Distinct from task-level errors in DATA_VALIDATION_ERROR. See Validation results.
Network rule¶
Snowflake object listing host and port pairs that a service may reach (MODE = EGRESS, TYPE = HOST_PORT). Used with an external access integration for SPCS Workers.
Normalization¶
SQL expressions applied during validation so source and target values are compared in a canonical form (for example consistent date or numeric formatting). Normalization makes values comparable; it does not declare which diffs are acceptable. Override built-in templates with validationCustomNormalizations at the workflow root. Distinct from accepted transformations. See Customizing normalization and metrics.
NOT_
L3 row validation results for rows present on only one side: NOT_FOUND_TARGET (the row exists on the source but is missing from the target) and NOT_FOUND_SOURCE (the row exists on the target but is missing from the source). See Validation levels and result codes.
Orchestrator¶
AIM DMV component that runs in Snowflake, creates workflows and tasks, loads staged migration data, and evaluates validation results.
Partition¶
A subset of a table’s rows processed as one unit during migration or validation. Boundaries are stored in PARTITION_METADATA.
PAT (Programmatic Access Token)¶
A programmatic Snowflake authentication method recommended for Orchestrator and Worker connections. Key-pair authentication is also suitable. Avoid interactive methods such as SSO because AIM DMV opens many short-lived connections. See Connecting to Snowflake with a PAT.
POSSIBLE_
Provisional L3 row result recorded when a table defines accepted transformations and row hashing finds a mismatch. Reconcile promotes genuine problems to MISMATCH or removes accepted-only rows. Should not remain on a completed workflow.
Query tag¶
Snowflake session parameter set on every Orchestrator and Worker query. Filter by TRY_PARSE_JSON(query_tag):DMVF_WORKFLOW_ID in QUERY_HISTORY.
Row fingerprinting¶
Default L3 technique: MD5 hashes of normalized row content per partition. Mismatches trigger cell drill-down.
SNOWCONVERT_
Default Snowflake database where AIM DMV stores workflow metadata, task queue state, migration progress, and validation results. See The SNOWCONVERT_AI database.
Snowpark Container Services (SPCS)¶
Snowflake feature for running containerized services in your account. See Deploying on Snowpark Container Services.
Sync strategy¶
Incremental migration mode: none (full load), watermark (track changes by a watermark column), or checksum (platform-dependent).
Task / task queue¶
Unit of work assigned to a Worker. Pending and in-progress tasks are stored in TASK_QUEUE.
Tolerance¶
Relative numeric threshold for L2 metrics comparison (comparisonConfiguration.tolerance, default 0.001). Does not apply to L3 cell comparison.
UNLOAD¶
Redshift extraction strategy that exports data through Redshift UNLOAD to S3, then loads into Snowflake.
Worker¶
AIM DMV component that connects to the source system and Snowflake, executes migration extraction or validation queries, and reports results to the Orchestrator.
Workflow¶
Configuration and runtime state for one migration or validation run. Identified by WORKFLOW_ID in SNOWCONVERT_AI.DATA_MIGRATION.WORKFLOW.
WRITE_
Teradata extraction strategies: WRITE_NOS exports through Teradata object storage integration; TPT uses Teradata Parallel Transporter.