The enhancements to cloud databases are expected to help in the development of AI-based and real-time applications.
Google Cloud has updated its fully managed distributed SQL database service Spanner to add graph processing capabilities, dubbed Spanner Graph.
The update is expected to help developers build applications around AI use cases, such as smarter recommendation systems and fraud detection, according to analysts.
Spanner Graph, according to Steven Dickens, chief technology advisor at The Futurum Group, will address the increasing demand for advanced data processing and analytics solutions in various industries across the AI stack.
“Graph databases are particularly useful for retrieval-augmented generation (RAG) because they excel at modeling and querying complex relationships between data points. This capability enhances the retrieval of relevant information in AI applications, improving the accuracy and relevance of generated outputs,” Dickens said.
The rise in demand for knowledge graphs, and the GraphRAG pattern — that can provide context, making the relationships between vector embeddings explicit — is another reason behind the addition of graph processing capabilities to Spanner, said Tony Baer, principal analyst at dbInsight.
Spanner is still “essentially a relational database management system (DBMS),” and records data internally, including graph data in the form of tables with rows and columns, explained IDC’s research vice president Carl Olofson. “The new graph capability enables users to add graphs to existing relational databases, and to use graph math to process table data.”
Olofson explained that this is why Spanner might not “really be in a position to challenge” specialty graph databases, such as Neo4j, OrientDB, TigerGraph, and Aerospike Graph, supporting pure graph deployments in terms of performance.
Multi-model databases vs specialty databases
Google’s move to add graph processing capabilities to Spanner is consistent with the trend of DBMS firms, who want to be seen as providing strategic database support in helping enterprises consolidate their database needs, add multi-model capabilities or capabilities that are offered by specialty databases, Olofson said.
One such example is the addition of vector capabilities by cloud database providers, such as Oracle, AWS, Microsoft, Google, and MongoDB. Previously, vector capabilities were just offered by specialty databases, such as Pinecone, Weaviate, and Milvus.
Similarly, Spanner’s graph processing capabilities, according to Nucleus Research senior analyst Alexander Wurm will allow Google to compete with the likes of Neo4j, Amazon Neptune, and Microsoft Azure Cosmos DB.
These graph capabilities will allow developers simplify data management tasks as they can now use a single platform to query both structured and connected data, reducing the complexity of managing multiple databases, Dickens explained, adding that this efficiency gain will aid business operations.
Alternatively, dbInsight’s principal analyst Tony Baer pointed out that Spanner Graph fills a gap in the Google database cloud portfolio as it was “conspicuously missing until now.”
“Google is probably starting with Spanner first as it wants to address horizontal scalability which has long been the Achilles Heel of graph databases,” Baer said.
Interoperability between GQL and SQL
Spanner Graph or Spanner in general, according to Google’s vice president of engineering for databases, Andi Gutmans, will now support the graph query language (GQL) and the database will also provide interoperability with SQL.
The interoperability with SQL, according to analysts, will allow Google to leverage the widespread familiarity and usage of SQL among developers and data professionals, without them needing to learn a new query language.
Last month, Google had updated Spanner with an option to maintain dual-region configuration, which it claims will make it easier for enterprises to comply with data residency norms across countries with more limited cloud support, and at the same time ensure availability.
Earlier this year, the company said it would be adding vector capabilities, such as approximate nearest neighbor search (ANN) and exact nearest neighbor search (KNN) to most of its cloud databases, including Spanner.
While ANN is used to optimize search, in other words, reduce latency for large datasets, KNN is used to return more specific or precise search results on smaller datasets.
Google also said it would be packaging Spanner in a new way, dubbed Spanner Editions, which would include the cloud database in Standard, Enterprise, and Enterprise Plus editions.
The three editions with varied pricing are expected to target more enterprises and offer more flexibility, it added.
Bigtable gets SQL support
In other updates to its cloud databases, Google has added SQL support to its NoSQL database service Bigtable, which includes more than 100 SQL functions.
“From KNN for developing generative AI applications and JSON manipulation for log processing, to using data sketches for real-time analytics, it’s now easier than ever to build real-time, high-performance applications,” Google’s Gutmans said, explaining the added SQL support to Bigtable.
The addition of SQL support, according to The Futurum Group’s Dickens, will help meet the evolving needs of developers and enterprises seeking to leverage Bigtable’s performance and scalability using familiar SQL queries.
“The timing reflects, competitive pressure from rival cloud database providers and the growing demand for real-time analytics as well as the need for scalable database solutions that can handle diverse data types and workloads,” Dickens said, adding that the SQL support positions Bigtable as a more competitive offering in the market, potentially attracting a broader user base and driving more usage of cloud services.
“We are at the point where hyperscalers are looking to achieve data gravity and move data to their cloud platforms to provide a bedrock for AI use cases.” Separately, the company said that it was making new Spark connectors that support PySpark, Scala and SparkSQL, generally available, in order to support data scientists building data pipelines and training large machine learning models on Bigtable.
