Lucata Pathfinder

Run deeper analytics on larger graphs than ever before possible using the Lucata Pathfinder next-generation computing platform. The Pathfinder enhances common graph databases or graph engines developed in-house by eliminating scaling issues. It accelerates multi-hop analytics on graph databases with 1 trillion nodes (scale 40) and beyond with no data pruning required. Lucata leverages patented Migrating Thread technology to treat even massive graph databases as unsharded instances, enabling fast analytics with no MapReduce-like processing by delivering 16x the performance using only 1/10th the power of a comparable x86 server-based system.

Pathfinder Powers a New Generation of Graph Analytics

Pathfinder-A can power open source, commercial, or custom-developed graph databases using one of the following approaches:

  • In the native programming architecture, calls to the Lucata algorithm library run natively, including Breadth-First Search (BFS), PageRank, Connected Component, Triangle Count, and K-truss Subgraph
  • API calls for GraphBLAS queries supported on the Lucata platform, such as Sparse Matrix Multiply. The open source LAGraph library provides an extensive collection of algorithms constructed with this hybrid programming architecture.

Access the Lucata SDK (Software Development Kit) for more details.

Pathfinder-A can easily scale to meet your needs with no manual reprogramming required; it can run analytics on petabyte-scale graph databases. A Lucata Pathfinder-A chassis has 512 GB of RAM, which allows you to run real-time analytics on a graph database with no sharding while storing up to 4 TBs of graph data in near-memory. Larger memory configurations will be available by the end of 2024. Each Pathfinder-A chassis can support up to eight 100 GBit I/O ports, allowing your graph database engine to ingest data at its maximum rate. A Pathfinder rack seamlessly connects eight Pathfinder chassis, providing up to 4 TB of RAM.  Over 1,000 racks can be interconnected as a single shared memory image, supporting over 4 PB of RAM.

There is no data pruning required for graph analytics or machine or deep learning training, eliminating the potential for introducing bias during pruning or sharding processes.


Pathfinder is a Smart Choice

The Pathfinder-A platform is cost-effective. Patented Lucata migratory thread approach to computing means compute threads move to your data rather than moving data to the CPUs. This approach results in much lower interconnect bandwidth requirements between Pathfinder chassis and racks than current approaches to distributed computing clusters require. Migratory thread computing allows Pathfinder chassis to consistently maintain high CPU and RAM utilization rates, lowering your operating costs.

Pathfinder-A is a new computing paradigm for big data that can process petabyte-scale datasets in real-time with no data pruning or database sharding. Initially available for graph databases, the Pathfinder-A platform has the capability to power massive performance gains for a variety of other database types. Pathfinder can also power high-performance machine and AI model training on sparse data.


Scalable Shared Memory

Unprecedented performance and scalable to 16PB of RAM for in-memory databases and data lake acceleration


Cache-less Design

Removes coherency cost and complexity overheads in shared memory databases for OLTP


Ultra Lightweight

Ultra lightweight computational elements for MPP databases combined with Migratory Threads for OLAP enable efficient queries without data movement


Low Power Consumption

Uses one-tenth the power of a comparable system built using conventional servers