Document database Couchbase is adding a columnar side-car to boost analytics performance for users who want more insight into their real-time data.
Announced at AWS re:Invent 2023 in Las Vegas, the new service introduces a columnar store and data integration into the Capella Database-as-a-Service (DBaaS) for applications such as customer profiling and special offers.
The in-memory analytics system is only available as a package with the main DBaaS, but offers support for Tableau and PowerBI for analytic development and visualization, the company said. It also launched a conversational coding tool dubbed Capella iQ designed to help developers use natural language interactions with ChatGPT for SQL++ development. Other LLMs will be added in the future.
The new service will be in private preview from next year and is expected to become generally available in fall.
Keen industry watchers may notice that MongoDB, a fellow document database designed for modern internet-native applications, added analytics features last year. The company created column store indexing to help developers create and maintain a purpose-built index to speed up many common analytical queries without requiring any changes to the document structure or having to move data to another system. Analysts said it might be a good system for straightforward queries but not complex modeling.
Couchbase emphasized the differences in its approach. It said MongoDB created a duplicative indexing structure against the data that persists in its singular storage engine, WiredTiger. Couchbase claimed WiredTiger consumes half the available memory when in use, “which is one of the reasons that MongoDB does not scale as efficiently as Couchbase and Capella.” The Capella approach means both columnar and document engines work and scale their workloads independently while living in the same cluster, the company claimed. We’ve asked MongoDB about this and will update if they respond.
Chris Bridgland, senior director, solutions engineering and customer success Europe, said the columnar store supported Avro files popular in the telco industry among other features allowing the system to analyze third-party data in its DBaaS.
He said another difference in the Couchbase approach was that it creates the index and the schema on the read, rather than before bringing in the data.
“We were already seeing in the current testing around about two to two and a half times improvement in performance,” Bridgland said.
Doug Henschen, vice president and principal analyst with Constellation Research, said the columnar database move would appeal to Couchbase customers looking for more analytical capabilities to go along with the platform’s transactional capabilities.
“The company has had analytical capabilities for five years and more than 30 percent of its customers are using them,” he said. “This announcement brings analytical performance to the next level as required by many next-generation applications that blend transactional and analytical needs. For example, a large number of Couchbase customers store tons of loyalty program data. Capella’s columnar and real-time capabilities could be used to power personalized offers and recommendations to customers in near real time.”
However, it was not necessarily a competitive gain against MongoDB, which has had analytical capabilities for several years and deepened them last year. “I don’t see it so much in a competitive context as I do in a case of two database providers helping customers to build next-generation applications including analytics on their respective platforms,” Henschen added.
He said it would be difficult to determine the differences in performance until customers started building proof-of-concept systems.
While Snowflake and SingleStore have both laid claim to performing analytics and transactions on the same system, the emergence of so-called “transanalytical” capabilities has been greatly exaggerated, Henschen said.
“There are niche databases that support both,” he said, “but they’re far from mainstream. And both Couchbase and MongoDB will be the first to tell you that their respective analytical capabilities will not displace or compete head-on with analytical data platforms like Snowflake or Databricks. What they’re after is providing analytical capabilities for the operational data within the applications built on their databases. I think we’ll continue to see separate, specialized products for some time to come.” ®
