Couchbase NoSQL database users can now get AI vector search across their edge sites and real-time analytics in the AWS cloud with Capella Columnar.
Couchbase provides its on-premises Enterprise Server database, which combines RDBMS and NoSQL analytics. It has expanded it into a database-as-a-service (DBaaS) offering, Capella, available on AWS, Azure, and Google Cloud Platform. A third offering is Couchbase Mobile, an integrated document database for mobile, desktop, and custom embedded hardware with sync capabilities built into Couchbase Enterprise Server and Capella. It is now introducing three new features:
- Capella Columnar on AWS enabling real-time analytics as well as operational workloads
- Couchbase Mobile with Vector Search to Support AI Extended Language Model (LLM) Inference Workloads
- Capella Free Tier for Developers
Matt McDonough, SVP of Products and Partners at Couchbase, said, “With the launch of Capella Columnar, we’re solving long-standing challenges in JSON data analytics, enabling businesses to seamlessly integrate insights into their operational applications. Our vector search capabilities in Couchbase Mobile extend this adaptability to Edge and IoT devices, opening up new possibilities for hyper-personalized and contextual applications.”
Vector search is already available in Couchbase Enterprise Server and Capella. It is used in AI generative inference by LLMs to create answers to user queries. Now, Couchbase Lite, the database for Couchbase Mobile, also integrates it, enabling semantic search and retrieval augmented generation (RAG) to provide more accurate answers.
McDonough said, “By enabling on-device vector search, we’re not only accelerating applications and eliminating downtime due to internet outages; we’re also opening up new possibilities for developing secure AI. Our solution combines the scalability of the cloud to handle massive AI datasets with industry-leading capabilities for immediate on-device processing. This unique combination, along with our support for multiple search techniques accessible through a single SQL++ query, enables developers to build sophisticated GenAI applications that can run entirely on-device.”
Couchbase Columnar has been compared to a massively parallel processing (MPP) transactional database within Capella, and analytical processing is kept separate from real-time operations.
Data can be loaded into Capella Columnar from AWS Dynamo, S3, MySQL, and other source datasets connected to Apache Kafka with Change Data Capture efficiently synchronizing updates. Columnar is supposed to “make JavaScript Object Notation (JSON) data useful in analytics.” The company said that “a lot of semi-structured JSON data sits dormant. Couchbase offers key-value and columnar storage options for both operational and analytical workloads on a single platform, giving customers the power and flexibility to make JSON data useful in analytics.”
Columnar makes it easy to “parse, transform, and preserve JSON data in a columnar format that’s ready for analysis. It supports real-time multi-source data ingestion, not only from Couchbase, but also using common systems like Confluent Cloud… to pull data from other third-party JSON or SQL systems.”
With Couchbase’s AI-powered Capella iQ coding assistant, which can write SQL++, once a metric is calculated, it can be immediately written back to the operations side of Capella, which can then use it in its analytics. Couchbase said this can be used to “embed customer-facing metrics into a gaming application to accelerate engagement.”
McDonough added: “This rewrite problem has gone unsolved by analytics systems for decades because it was too difficult to anticipate what a developer would do with it. Capella Columnar implements the solution, and the needs of AI-based applications provide the motivation.”
We expect Columnar support to be extended to Capella running in Azure and GCP clouds.
Carl Olofson, research vice president at IDC, said Couchbase has “added support for mixed analytics and transaction processing that leverages the performance advantage of columnar data management as well as vector search to support applications that require intelligent access to data… These are capabilities the market has been looking for but are hard to find in a single product.”
Couchbase competitor SingleStore also supports vector searchCouchbase and SingleStore both compete with dedicated vector database vendors such as Pinecone, Qdrant, Milvus, and Zilliz. They claim that vector database technology is a product, while Couchbase and SingleStore claim that it is a feature.
Capella Columnar is available in a standard version, as is Couchbase Mobile with vector search. The free tier of Capella will be available in September.
Read a Couchbase announcement blog here. Watch a video on Couchbase Mobile with vector search and learn more here.
