Enterprise AI infrastructure and graph database company TigerGraph has moved on to pastures new with Savanna, its cloud-native data platform offering designed to enable data engineering teams to discover relationships and inter-relationships between data sets, to analyze complex patterns and make decisions across large amounts of highly connected data. The company wants to position graph database technology as a supercharge for AI systems.
What is graph database parallelism?
TigerGraph details its Native Parallel Graph (NPG) design, where the “parallelism” aspect is a dual and concurrent focus upon both storage and computation. This technology supports real-time graph updates and built-in parallel computation.
According to magical analyst house Gartner, graph analytics are also a critical enabler of natural language processing and data fabrics by virtue of their ability to help enrich and accelerate data preparation, analytics and AI. The wizardly quadrangle-based technology soothsayers at Gartner also note that graph analytics processing is core in other advanced technologies including virtual personal assistants and smart advisors i.e. such as those found in so-called “robo-advisors” which can be used to guide users in passive financial investing practices.
The company says that TigerGraph Savanna enables NPG massively parallel storage and computation to scale independently and provision (what is claimed to be) six times faster than alternative graph database offerings. Data engineering and software application development teams using these services can also make use of a fully managed service to handle underlying infrastructure and maintenance. At the coalface, that means access via a user-friendly interface with options to connect to 300+ APIs for tasks such as deployment configuration and monitoring.
Pre-configured solution kits
“We already offer the fastest, most scalable graph database and analytics platform,” said Rajeev Shrivastava, CEO of TigerGraph. “Our Savanna cloud-native platform accelerates setup speed six-fold while providing agility in how companies consume storage, compute and managed service options. Nine new pre-configured solution kits encapsulating TigerGraph’s expertise for mission-critical uses including transaction fraud, application fraud, product recommendation, mule account detection, entity resolution, supply chain management and network Infrastructure.
Enterprises can choose between a fully managed service or deploying TigerGraph on their own cloud infrastructure, to best suit their operational and compliance needs. The company has stated that Savanna provides separate scaling of storage and compute, scheduled expansion/shrink and auto stop/resume features for enhanced elasticity. All that happens while the platform itself is able to maintain TigerGraph’s distributed, massively parallel graph database architecture.
OLTP OLAP overlap
Depending on the use case at hand, teams can optimise performance and management of online transactional processing (OLTP) and online analytical processing (OLAP) traffic through dedicated compute workspaces tailored for each workload. Deeper into the product specs here, TigerGraph has also opened up to data lake and cloud relational database management system (RDBMS) sources including Snowflake, Spark, Delta Lake, Iceberg and Postgres, in addition to AWS, Azure, and Google Cloud object stores.
In terms of support for query languages, users can choose from GSQL, OpenCypher, or the pattern-matching syntax of GQL, the new ISO standard graph query language, which TigerGraph helped to develop from GQL’s inception.
Do we need graph for AI?
All said and done then, do we need graph database technology to come to our rescue as some kind of “new enabler and accelerator” for the next phase of AI? Certainly, graph databases have a flexible schema that enables data engineers to create robust linkages and attributes to be associated with nodes and relationships. This has been argued to make them a better data source for AI than relational databases.
Fabien Vives reminds us that most users don’t need (or perhaps want) to become AI or graph experts. “They just want to get insight in the most effective way, understand where it came from, and make an informed decision,” explained Vives in his role as senior director of product management at e-merchandising platform company nfinite writing on Cambridge Intelligence.
He further suggests that graph visualization is the “natural way” to present a lot of this information, revealing connections, relationships and sequences. He enthuses that there’s a great visual attraction here too i.e. users are often are drawn to graphs for their intuitive and interactive format. Overall, graph databases have been lauded as excellent recommendation systems due to their ability to traverse relationships between users, products and preferences… and in the world of AI, quick connected thought is often a good thought.
Graph is good for AI they say, let’s chart its course.
Free (top) image: Wikimedia Commons
