Rockset

Rockset was a serverless, cloud-native real-time analytics database founded in 2016 by former Facebook engineers Venkat Venkataramani and Dhruba Borthakur.

Reviewed by 7wData

On this page

Publisher review

Rockset was a serverless, cloud-native real-time analytics database founded in 2016 by former Facebook engineers Venkat Venkataramani and Dhruba Borthakur. The platform delivered fully-indexed SQL analytics on schemaless JSON data without requiring upfront data modeling, using proprietary Converged Indexing technology that combined row, columnar, and document indices. Rockset ingested streaming data from sources like Kafka, DynamoDB, and relational databases, making it available for millisecond-latency SQL queries across complex joins and aggregations.

The platform targeted enterprise customers needing real-time operational analytics—fraud detection, anomaly monitoring, user-facing dashboards fed by live event streams. It operated as a fully managed SaaS on AWS, Google Cloud, and Azure, with a serverless consumption model: users paid for data ingestion, query volume, and storage, with entry-level deployments starting around $232 per month. Rockset gained significant traction among data teams building modern, event-driven analytics pipelines.

However, in June 2024, OpenAI acquired Rockset for an estimated $200–$300 million, acquiring the company primarily for its real-time indexing and retrieval capabilities to enhance OpenAI's infrastructure, not for its customer base. The acquisition led to Rockset's immediate discontinuation: existing customers received notice they had until September 30, 2024, to migrate, and the company deleted all customer data upon shutdown. The acquisition highlighted both Rockset's technical strength—severe enough to attract one of AI's largest players—and a structural vulnerability: the high acquisition risk for standalone infrastructure startups. Today, customers have migrated to alternatives including ClickHouse/Tinybird, Pinot/StarTree, Materialize, SingleStore, and Firebolt, each with different trade-offs in latency, query complexity, and operational complexity.

Get the AI & data signal, daily.

335k+ subscribers read this every morning. One email, both newsletters. Unsubscribe anytime.

How it works

  1. Converged Indexing

    Proprietary multi-index technology combining row, columnar, and document indices to deliver sub-second query latency on both structured and schemaless data without pre-defining schemas.

  2. Real-Time Data Ingestion

    Streamlined ingestion from Kafka, DynamoDB, relational databases, and cloud storage, with data immediately available for SQL queries rather than staged through ETL pipelines.

  3. Serverless Architecture

    Fully managed SaaS with automatic scaling and no infrastructure provisioning required; customers pay only for data ingested, queries executed, and storage consumed.

  4. Complex SQL Analytics

    Support for complex queries including multi-way joins, aggregations, and window functions across massive datasets with consistent sub-100ms latencies.

  5. No Schema Pre-Definition

    Schemaless JSON ingestion and query model eliminated upfront data structure definition, reducing time to analytics and accommodating rapid data model changes.

  6. Private Deployments

    VPC-isolated deployments on AWS, Google Cloud, and Azure ensured data never left customer infrastructure while remaining fully serverless.

Strengths and trade-offs

Strengths

  • Sub-100ms query latency on real-time data without manual tuning; customers reported excellent support (rated 9.4) for onboarding and troubleshooting.
  • No schema design overhead; ingest raw JSON and query immediately, reducing time from data source to insight for event-driven use cases.
  • Purpose-built for operational analytics: strong fit for fraud detection, anomaly flagging, and live dashboards where freshness and latency are critical.

Trade-offs

  • Vendor lock-in and acquisition risk: the OpenAI acquisition proved fatal for the product, stranding all customers on a 90-day migration deadline and destroying trust in startup analytics infrastructure.
  • Setup complexity: some users reported the initial deployment and configuration were more involved than competitors, and the service-level integration required closer partnership with Rockset support.
  • Higher cost at scale: tiered pricing (starting at $0.25/GB-month for small deployments, down to $0.13/GB-month at 100+ TB) could exceed purpose-built alternatives like Pinot or ClickHouse for very large datasets, especially under variable query loads.

Pricing context

Rockset used consumption-based pricing across three dimensions: data ingestion, query execution, and storage. Storage pricing was tiered: $0.25/GB-month for 0–100 GB, $0.20/GB-month for 100 GB–1 TB, $0.16/GB-month for 1–100 TB, and $0.13/GB-month for 100+ TB. Compute (virtual instances) and query volume added separate charges.

A free tier with limited resources was available for development. Entry-level production workloads with modest ingestion and query volume started around $232/month. The pricing model was transparent and allowed cost control via resource limits, but lacked the simplicity of single-price-per-feature competitors. All pricing is now obsolete as Rockset ceased commercial operations in September 2024.

Alternatives

User reviews

No user reviews yet. Be the first to write one.

Sources

Reporting on this tool draws on these publicly available sources.

  1. www.globenewswire.com — Rockset's January 2024 cost optimization (30% compute reduction, entry pricing $232/month) and positioning as serverless cloud analytics database.
  2. www.cbtnuggets.com — OpenAI's June 2024 acquisition rationale (data indexing and retrieval capabilities for AI applications), Rockset's core features (real-time analytics, complex queries, data compression), use cases (enterprise operational analytics), and 90-day migration deadline.
  3. medium.com — OpenAI acquisition announcement, impact on customers (forced migration), and Rockset's market position before shutdown.
  4. estuary.dev — Rockset's strengths (real-time analytics), post-acquisition alternatives (ClickHouse, Tinybird, Druid, Pinot, SingleStore, RisingWave), and the vendor instability risk that materialized.
  5. blog.dataengineerthings.org — Analysis of why OpenAI acquired Rockset and implications for the data analytics market.
  6. www.datanami.com — Rockset's SQL querying capability on DynamoDB and other NoSQL sources.
  7. www.infoq.com — Technical architecture of Rockset: RocksDB foundation, Converged Indexing, distributed design.