Rockset takes a deeper dive into enterprise data pool
Coming off its last venture round last fall, Rockset is now training its sights on the more traditional data warehousing market with support for MySQL and PostgreSQL sources. It makes for an updated take on the traditional symbiotic relationship between transaction databases and data warehouses.
To recap, Rockset is a managed database cloud service that capitalizes on RocksDB, the open source project from Facebook that is a compact, key/value storage engine written in C++. The company founders created RocksDB.
Rockset is a classic open core company whose secret sauce is a converged index that makes the data look relational and queryable by SQL. Compared to platforms like Snowflake (one of its biggest sources of customers), Rockset doesn't compress data, but it uses compute far more efficiently. So that puts the sweet spot for Rockset deployments in the multi-terabyte range. Rockset claims that its high performance engine provides a lower cost alternative to cloud data warehouses like Snowflake. Rockset has also drawn customers from DynamoDB and MongoDB seeking to perform analytics.
By adding MySQL and PostgreSQL support, Rockset is taking on a familiar role in the relational database world; providing the data warehouse that offloads analytic queries from the transaction database. But there are some key differences. Whereas traditional data warehouses involved ETL processes (usually in batch) to summarize transaction data to reduce its storage footprint, Rockset takes advantage of the economics of cloud storage to store the data in original form. In place of database extraction processes, it takes continuous change-data-capture (CDC) feeds from database logs and is updated almost in real-time, with any delay attributable to the time it takes to transit the network.
In turn, Rockset exposes data via SDKs and integrations with visualization tools such as Tableau, Retool, Redash, and Superset. With MySQL and PostgreSQL support now live, we expect that Rockset will next train its sights to more of the usual suspects in enterprise databases.