A Breakthrough Approach to Addressing the Toughest Big Data Challenges
Lucata Pathfinder systems are built from the ground up for unprecedented performance in data intensive big data environments. The system delivers a truly scalable shared memory which simplifies programming as compared to clusters with MPI or GPUs. The cache-less design overcomes previous system generations’ shared memory inhibitors which include coherency cost and complexity. And, unlike other computers, our migrating thread technology dramatically reduces data movement across the network by moving just the program context rather than large data chunks. Our solutions leverage familiar, industry standard software and are straight-forward to query and program. We do this while building energy efficient technology solutions that dramatically reduce data movement which results in reduced power consumption. This enables breakthrough capabilities for NoSQL, NewSQL, graph query and data science focused databases with integrated machine learning
Scalable Modular System
The Lucata Pathfinder is a modular, strong-scalable system that enables users to optimally size and grow their analytics and database systems as needed. A single rack mountable chassis, containing 8 nodes, can simultaneously execute over 12,000 concurrent threads. The Lucata Compute Elements are specifically designed to deliver scale and performance on high volume, small transactions and queries. The system utilizes a unique cache-less, narrow channel memory access system that enables 8-byte transfers foundational to the efficient implementation of migratory threads. Racks can be combined within a 100M radius to create a massive single image of unified memory coupled with millions of parallel threads and hundreds of thousands of PCIe slots for parallel ingest, streaming I/O and storage expansion.
Linux Operating Environment and Ecosystem
Each node board includes a standard processor for Linux and service requests, along with multiple Lucata Compute Elements that handles massively parallel compute utilizing migrating program contexts (threads) to dramatically reduce data movement. The Pathfinder is programmed in Cilk, which is an enhancement of C/C++ that adds three keywords for parallel asynchronous threading. An open API enables industry standard application software to run on the standard processor while calling optimized algorithms and parallel codes on the Lucata Compute Elements, opening a robust ecosystem of tools that enables users to work in familiar environments while tapping into Lucata’s innovative architecture. The Pathfinder system includes a massively parallel data-base storage engine to accelerate SQL and graph queries at scale. GraphBlas and RedisGraph have been ported to the Lucata Platform to exploit the architecture.