This article is more than 1 year old

MongoDB and Rockset link arms to figure out SQL-to-NoSQL application integration

NoSQL, no problem for Facebook-originating RocksDB

MongoDB and fellow database biz Rockset have integrated products in a bid to make it easier to work with the NoSQL database through standard relational database query language SQL.

The joint development is designed to make MongoDB – a NoSQL, document-oriented database – much more painless for developers to interrogate in applications that perform SQL Joins, for example.

This has, so far, been difficult: Essentially, schema-less MongoDB does not support SQL Joins, a way of combining records of tables using SQL, although there are ways to get around the problem.

The Rockset-MongoDB integration claims to solve Joins problem.

"Rockset replicates MongoDB data and builds an external index on that data in real time," a Rockset spokeswoman told The Register.

"We use a modern approach to building our index that we call Converged Indexing which combines the benefits of a search index and a columnar index in a single system. This indexing approach has two remarkable properties – it requires no configuration ahead of time and accelerates a wide spectrum of data retrieval patterns out of the box. We've also built a new distributed SQL query engine along with a query optimizer that exploits the speed ups delivered by converged indexing."

MongoDB gained popularity in the mid-2010s within web applications by helping to provide personalised content without developers worrying too much about database schema. It is used in mobile, web, gaming, IoT and other applications by companies including PayPal, Google and Facebook.

RocksDB was originally developed by Facebook engineering, which open-sourced it to the broader community. Rockset co-founder and CTO Dhruba Borthakur was one of the founding engineers of RocksDB.

Rockset said the product integration is designed to make building applications such as recommendation engines faster and easier.

The mouthpiece added: "Building this will involve three types of data sets: offline recommendations generated by nightly machine learning jobs that usually sit in a data lake, past order history for the customer that is tracked in MongoDB, and recent behavioural data to find the products the customer looked at from clickstream logs. With Rockset, data from each of these three data sources can be seamlessly loaded and indexed in real-time for fast SQL processing without any Extract-Transform-Load (ETL) jobs or data pipelines." ®

More about


Send us news

Other stories you might like