SYSTEM Cited by 1 source
StarRocks¶
StarRocks is an open-source, MPP (massively parallel processing) OLAP analytical database engineered for real-time and high- concurrency analytical workloads. It is a common choice alongside Apache Druid for the OLAP axis of workload-specific storage migrations.
Definition¶
StarRocks features:
- Columnar storage with vectorised query execution.
- MPP architecture — a coordinator splits work across backends that each scan their local shards in parallel.
- Materialised views with automatic rewrite.
- High concurrency — tuned for dashboards and customer-facing analytics, not just BI batch queries.
- Real-time ingestion from Kafka, local files, or direct writes.
Historically StarRocks grew out of the Apache Doris project; it is developed by StarRocks Inc. with a strong open-source core.
Role in Pinterest's HBase migration¶
StarRocks is named alongside Druid as one of the two OLAP-axis successors to HBase at Pinterest:
"online analytics workloads would be migrated to Druid/StarRocks, time series data to Goku, [...] and key value use cases to KVStore." (Source: sources/2024-05-14-pinterest-hbase-deprecation-at-pinterest)
HBase "performed worse than state-of-the-art solutions for OLAP workloads." Columnar stores like StarRocks beat wide-column NoSQL on analytical scans by 10×+ because of storage layout + vectorised execution — canonical patterns/workload-specific-datastore-migration rehoming of the OLAP axis.
Seen in¶
- sources/2024-05-14-pinterest-hbase-deprecation-at-pinterest — named as an OLAP-axis successor to HBase for Pinterest's online-analytics workloads.
Related¶
- systems/apache-druid — sibling OLAP substrate named alongside StarRocks in Pinterest's migration.
- systems/hbase — the store being replaced.
- concepts/olap — the workload category.
- patterns/workload-specific-datastore-migration — the decomposition pattern.
- companies/pinterest.