```json { "table_name": "query_log", "primary_key": "query_id", "candidates": [ { "rank": 0, "partition_key": "dataset_id", "partition_key_label": "E", "partition_key_reasoning": "RPUV=30.65 (83 distinct datasets across 2,544 queries) is ideal for partitioning. Top 3 coverage is 9.63%, indicating good distribution. Semantically, dataset_id is the entity whose query activity we trace — each dataset accumulates a timeline of queries run against it.", "clustering_key": "created_at", "clustering_key_label": "S", "clustering_key_reasoning": "created_at is the event timestamp with perfect uniqueness (1.000) and natural temporal ordering. Clustering by created_at within each dataset_id groups all queries for a dataset in chronological order, enabling efficient time-range scans and query history analysis.", "world_line": "dataset_id", "session_column": "batch_id", "confidence": "HIGH", "notes": "Primary perspective: queries grouped by dataset. Each dataset has a timeline of analytical queries. batch_id (RPUV=24.46) can serve as a session identifier grouping related queries within a single analysis run." }, { "rank": 1, "partition_key": "tenant_id", "partition_key_label": "E", "partition_key_reasoning": "RPUV=2544 (single value 'default') makes this a degenerate partition — all rows in one file. While semantically tenant_id is the top-level entity, it offers no distribution benefit in this dataset. Included as an alternative only because multi-tenant systems may eventually have multiple tenants.", "clustering_key": "created_at", "clustering_key_label": "S", "clustering_key_reasoning": "Same as rank=0: created_at orders all queries chronologically within the single tenant.", "world_line": "tenant_id", "session_column": "batch_id", "confidence": "LOW", "notes": "Alternative perspective: queries grouped by tenant. Currently all data belongs to 'default' tenant, so this degenerates to a single partition. Only viable if the system scales to multiple tenants in the future." } ] } ```