Katana Graph Inc., developer of a platform that analyzes large unstructured data sets using high-performance graph algorithms, announced today that it has raised $28.5 million in a new round of funding.

The Series A financing, led by Intel Capital, will go toward challenging some well-heeled competitors in the hot graph database market.

Katana’s engine, which was developed over the past decade by scientists at the University of Texas at Austin, is billed as more of an analytics engine than a database management system. In fact, the company is still in the process of building a database that Chief Executive Keshav Pingali said will be “a full-stack solution going from storage and ingest to in-memory graph to graph [artificial intelligence] and analytics.”

For now, the startup is focusing on its analytics capability, which it says can operate on very large data sets with the ability to make sense of unstructured data. “Sparse data sets can usually be looked at as graphs and we can bring structure to them,” Pingali said, citing the example of large social networks in which relationships don’t lend themselves to a rigid row-and-column structure.

One unnamed customer has built a model incorporating 4.3 trillion edges, which are the connections between nodes in a graph network. “We can handle that and nobody else can,” Pingali said.

Scalable engine

Katana Graph says its engine outperforms all others and can scale up to clusters of 256 machines, runtimes and communication paths optimized for graph analytics. The engine supports heterogeneous clusters of computing resources including x86 central processing units, Arm Ltd. CPUs, graphics processing units and other accelerators.

The company touts ease of development for the application of graph mathematics to AI, graph pattern mining and graph analytics. Katana Graph works with the popular Python data science programming language and supports OpenCypher, an open query language for property graph databases. Support for the Graph Query Language standard will be added when that specification is finalized.

The technology was honed over the course of several projects conducted for the Defense Advanced Research Projects Agency and has been used by Intel Corp. to optimize design of its Xeon chips and GPUs, Pingali said. Other customers include a large pharmaceutical company that’s building medical knowledge graphs and a financial technology customer doing graph pattern mining for security and identity management.

The lack of a native graph database hasn’t been an impediment, Pingali said. “Most customers are not using graph databases but CSV files, NoSQL stores and so on, so we have built specialized adapters to read that data into memory,” he said. Pingali claimed that the company’s ingestion engine has been able to reduce data load times from “a couple of days to 20 minutes.”

The funding comes just a week after TigerGraph Inc. pulled in $105 million in venture funding, vaulting it to the top ranks of the emerging graph database industry. Pingali praised TigerGraph and market leader Neo4j Inc. for having “done a wonderful job of building systems you can query.” He said his company expects to distinguish itself by its focus on analytics.

Other participants in the funding round include Walden International Inc., Nepenthe LLC, Dell Technologies Capital and Redline Capital Management SA. Pingali said the funding should support operations for 12 to 18 months and enable further expansion of offices in Austin, the Bay Area and New York City, with a Seattle presence planned as well.

Photo: Unsplash