Katana Graph Accelerates Large‑Scale Graph Analytics on 3rd‑Gen Intel Xeon with Double Speed Boost

Katana Graph, a high‑performance graph analytics startup, has announced that its flagship Katana Graph Engine (KGE) has been fully optimized for Intel’s 3rd‑generation Xeon scalable processors and the accompanying Optane persistent memory platform. The upgrade delivers a near‑double speed increase for graph workloads on the new hardware while enabling larger datasets to be processed on smaller clusters.

The KGE is a distributed runtime that blends CPUs and GPUs into a single, cohesive platform. It supports advanced graph analytics, pattern mining, and querying on massive, irregular data sets—including web‑crawl graphs with billions of vertices and trillions of edges. By leveraging both horizontal scaling and high‑bandwidth memory, the engine can analyze graphs that would be infeasible on a single machine.

Developed over a decade of research by co‑founders Chris Rossbach and Keshav Pingali, the engine incorporates cutting‑edge techniques in graph partitioning, NUMA‑aware load balancing, and GPU kernel optimization. The partitioner can be customized with user‑defined policies, while the communication runtime minimizes the overhead that typically plagues distributed graph algorithms.

For Python developers, the KGE exposes highly scalable concurrent data structures—such as thread‑safe graph representations and worklists—allowing rapid prototyping without sacrificing performance. The runtime automatically handles data distribution, memory management, and cross‑node coordination, freeing data scientists to focus on algorithm design.

With the 3rd‑gen Xeon scalable processors, Katana Graph’s platform takes advantage of the latest micro‑architectural improvements, including higher core counts, larger cache hierarchies, and Intel Deep Learning Boost. When paired with Optane persistent memory, the system offers a dramatic increase in effective DRAM capacity, further reducing the need for expensive, large clusters.

“Our engine now delivers up to twice the performance on 3rd‑gen Intel Xeon than on the previous generation, right out of the box,” said Keshav Pingali, co‑founder and CEO. “By working closely with Intel, we’ve unlocked new levels of efficiency that our enterprise customers—spanning retail, finance, and identity management—can directly feel in their day‑to‑day analytics.”

Intel’s Vice President and General Manager of Machine Learning Performance, Wei Li, echoed the enthusiasm: “Katana Graph has harnessed the innovative architecture of the 3rd‑gen Xeon scalable processor, delivering a game‑changing boost in graph analytics performance. Our collaboration will help customers tackle the biggest unstructured data challenges.”

Beyond raw speed, the new optimization reduces cluster footprint and operational costs. Katana Graph’s mission is to give organizations the fastest, most cost‑effective graph software stack, empowering them to extract actionable insights from unstructured data and accelerate business growth.

In a recent milestone, Katana Graph raised $28.5 million in a Series A round led by Intel Capital, with participation from WRVI Capital, Nepenthe Capital, Dell Technologies Capital, and Redline Capital. The funding underscores the market’s appetite for high‑performance graph solutions across sectors such as pharmaceuticals, fintech, identity, and cybersecurity.

Founded in 2020 by professors from the University of Texas at Austin, the company combines academic rigor with commercial expertise. Its leadership roster includes Farshid Sabet, former GM of Edge AI at Intel, and board advisors such as Infosys founder N. R. Narayana Murthy and Cisco‑P&G executive Amy Chang.

Katana Graph’s unique value proposition lies in its ability to handle massive unstructured data, support complex pattern mining, and deliver superior performance for graph AI and analytics on heterogeneous clusters—including x86 CPUs, Arm CPUs, GPUs, and other accelerators.