Apache Pinot vs StarRocks

Compare and contrast Apache Pinot and StarRocks by architecture, ingestion, queries, performance, and scalability.

Compare Apache Pinot to Rockset here

Compare StarRocks to Rockset here

Apache Pinot vs StarRocks Architecture

Architecture

Apache Pinot

StarRocks

Deployment model

PaaS or self managed

Use of storage hierarchy

Hot storage plus Deep Store for backup and restore operations

Data is stored on disk and in memory

Isolation of ingest and query

No, but you can limit resources for ingestion and querying separately

Separation of compute and storage

No, although StarTree separates compute and cloud storage, which is an important consideration for those considering StarTree vs. open-source Pinot

No, but StarRocks supports nodes that don't store data locally

Isolation for multiple applications

Full isolation with replication

Pinot is a real-time distributed OLAP datastore that ingests both batch and streaming data. It has a distributed systems architecture that scales both horizontally and vertically, but unlike alternative OLAP databases, it does not decouple storage and compute. It supports both self-managed and PaaS options.

StarRocks is a high-performance OLAP database that can be deployed on the cloud or self managed. StarRocks does not separate compute and storage and offers limited options for resource isolation. It offers a robust set of features and high performance but requires considerable expertise to operate and scale.

Apache Pinot vs StarRocks Ingestion

Ingestion

Apache Pinot

StarRocks

Data sources

Streaming • AWS Kinesis • Apache Kafka Batch • Cloud Storage • PostgreSQL • MySQL • Snowflake • Google BigQuery File upload Write API

Streaming • Kafka • Flink Data lakes • HDFS compatible • Cloud storage

Semi structured data

Support for JSON column types, but JSON must first be converted to a string

Supports columns with JSON data • Does not support mixed-type columns • Support for star and snowflake schemas

Transformations and rollups

Yes

Yes, via materialized views

Pinot supports high-performance ingest from streaming data sources. Each table is either offline or real time. Real-time tables have a smaller retention period and scale based on ingestion rate while offline tables have a larger retention period and scale based on the amount of data. In order to persistently store the generated segments that make up a table, you will need to change controller and server configs to add deep storage.

StarRocks ingests data from a variety of sources, including both batch and streaming data. StarRocks can ingest nested JSON data, but enforces type at the column level.

Apache Pinot vs StarRocks Performance

Performance

Apache Pinot

StarRocks

Updates

• By default, all data in Pinot is immutable • Upserts only supported for streaming ingest • No support for upserts on data using the star tree index • No support for upserts on out-of-order events

While StarRocks is mutable, the update rate is slow, which is why it is most often used for append-only workloads

Indexing

• Manually configured • Forward index, inverted Index, Star-Tree index (columnar), bloom filter, range index, search index, JSON index, geospatial index, timestamp index

Columnar index, limited support for inverted indexes

Query latency

50-1000ms queries on 100s of TB

Storage format

Pinot stores data in a columnar format and adds additional indexes to perform fast filtering, aggregation and group by

• StarRocks is a columnstore that organizes data into prefix indexes, per-column data blocks, and per-column indexes • All data is replicated 3 times to achieve both fault-tolerance and concurrency

Streaming ingest

1-2 second ingest for streaming data

Data latency is typically 1-2 seconds

Like its competitors, Pinot can achieve sub-second query latency at high concurrency. However, this level of performance requires tuning, management, and deep expertise. Compared with the open-source version, the PaaS versions of Pinot address some of these issues, but similarly require expertise while making tradeoffs affecting query performance.

StarRocks was purpose-built for high-performance ingest, low-latency queries, and high concurrency. Optimized performance requires significant manual tuning.

Apache Pinot vs StarRocks Queries

Queries

Apache Pinot

StarRocks

Joins

SQL joins limited to inner join, left-outer, and semi-join

Multi-table join support

Query language

SQL

Developer tooling

• API for querying SQL via POST command • Clients for JDBC, Java, Python, and Golang • Integrations with Trino and Presto

Minimal

Visualization tools

Integrations with Tableau and Superset

Compatibility with MySQL protocols enables StarRocks to work with BI tools

In Pinot, SQL-like queries are received by brokers and scatter the request between real-time and offline servers. The two tables then process requests, send results back to the broker, and responds with the result. Joins are limited, as is support for UDFs and subqueries, making Pinot more or less useful depending on the use case.

StarRocks uses a high-performance vectorized SQL engine, a custom-built cost-based optimizer, and has support for materialized views.

Apache Pinot vs StarRocks Scalability

Scalability

Apache Pinot

StarRocks

Vertical scaling

All 4 pinot node types can be resized manually

• Both frontend and backend nodes can be manually resized

Horizontal scaling

• Users can add additional Pinot nodes to scale horizontally • Rebalancing is manual

• Both frontend and backend nodes can be manually scaled horizontally

Pinot allows for vertical scaling by increasing CPU and memory for each node as well as horizontal scaling by adding additional nodes. Capacity planning is a time-consuming, iterative, and manual task. It involves load testing and tuning across multiple vectors including read QPS, write QPS, number of streaming partitions, daily data size, retention period, types of workloads, number and type of segments, and much more.

StarRocks can scale up or out, but its tightly coupled compute and storage scale together for performance. This often results in resource contention and overprovisioning. Scaling StarRocks often requires deep expertise as there are many levels of the system that need to be managed.