SYSTAP: Scalable, High-Performance Graph Database, and Analytics

Brad Bebee, CEO & Managing Partner
In today’s connected world, organizations ranging from banks to large government agencies need to manage and mine large amounts of data. There is a constant need for analytics at an ever-increasing data scale. Organizations are in need of scalable solutions such as Big Data-driven graph analytics to address the challenges of analyzing large amounts of data in seconds. “Graph and machine learning analytics enable organizations to deliver better ROI outcomes than ever before by delivering new insights from their linked data,” says Brad Bebee, CEO and Managing Partner, SYSTAP. The Greensboro, NC-based company was founded with a vision of building high quality, highly scalable software solutions for big graphs.

Over the years, SYSTAP has offered a set of highly scalable software solutions to help scale graphs. SYSTAP’s flagship product, Blazegraph, is a high-performance graph database that features robust, scalable, and enterprise-class storage with online backup, high-availability and scale-out. In 2015, SYSTAP discovered a way to exploit the main memory bandwidth advantages of Graphics Processing Units (GPUs) to deliver graph analytics 100s of times faster than CPU-based approaches. These advances are being integrated into the Blazegraph family as Blazegraph GPU and Blazegraph DASL (pronounced “dazzle”).

Blazegraph’s embedded and single- server deployment modes offer a scalable graph database platform for supporting the RDF/SPARQL and Tinkerpop/Blueprints graph APIs. A single server can support and scale data up to 50 billion edges using server-class machines. It has a REST API to support query, data loading, and database management functions. Besides, both the embedded and single-server modes support stored queries for custom application logic, custom services and indices, custom functions, and vertex-centric programs. In late 2015, SYSTAP has launched Blazegraph GPU to provide drop-in GPU acceleration for its core Blazegraph product. “With Blazegraph GPU, you get 200-300X graph query acceleration just by switching to a GPU-enabled platform with almost zero changes to your application,” says Bebee.

Our Blazegraph products enable customers scale their graph databases and analytics with solutions ranging from embedded applications to GPU clusters

In addition, SYSTAP’s Blazegraph’s DASL (“dazzle”) helps organizations process large graphs in near real time. DASL can run on both one andmultiple GPUs—a cluster of 64 NVIDIA K40 GPUs—to traverse a scale-free graph of more thanfour billion directed edges in milli-seconds. This helps in achieving a throughput of 32 billion traversed edges per second (32 GTEPS). Blazegraph DASL provides a Scala-based language to write graph analytics and in complementary to the Spark and Hadoop data ecosystems. Users of DASL typically see a 1000X increase in their graph analytics versus running directly in Spark.

SYSTAP’s software is in use by Fortune 500 customers such as EMC, Autodesk, and many others for high performance graph database applications and analytics. For Instance, Wikidata, a central storage for all Wikimedia Foundation data—selected SYSTAP’s Blazegraph to be the graph database platform for the WikiData Query Service—a new capability to allow users curate the content contained in Wikidata.

With its distinct expertise in the Big Data landscape, SYSTAP is a go to company for scaling graphs. Presently, SYSTAP is focused on bringing out the full potential of GPUs for enhancing graph analytics operations. The company is planning to further develop their products by incorporating new software and hardware technologies by next year. “Our GPU technology makes high performance graph analytics 40X more affordable than competing solutions. We strongly believe 2016 will be the year of the Blazegraph for big data graph analytics,”concludes Bebee.


Greensboro, NC

Brad Bebee, CEO & Managing Partner

Provider of high scalable software solutions for solving complex graph and machine learning algorithms