Custom Supersystems

For the past decade, we've been the leading innovator in large-scale Linux Supercomputing, from delivering the first ever Linux Supercomputing cluster to deploying multiple thousand-plus node production compute systems. Custom Supersystems are architectured and tuned to meet the most challenging, next generation needs of our customers – including applications requiring extreme scale or those benefiting from emerging technologies. Custom Supersystems achieve new levels of Total Application Throughput by blending best-of-breed supercomputing components within tightly-integrated, highly-tuned supercomputing architectures. Each Custom Supersystem is designed by Linux Networx architects and application experts to meet the unique performance, scalability and reliability requirements of leading edge customers.

Linux Networx continually evaluates a range of emerging technologies to select and validate next generation performance supercomputing components. Only those technologies that meet our stringent standards are leveraged within our Custom Supersystem practice. Supercomputing technologies are evaluated based on:

• Their peak and sustained performance in scalable computer system environments
• Reliability, manageability, interoperability
• Overall price/performance value
• After deployment, technologies are re-evaluated based on real-world results to assure success

Linux Networx Custom Supersystems are specifically designed to deliver:

Extreme supercomputing performance – Large-scale compute and visualization systems achieve tens to hundreds of Teraflops and total graphics performance of over 45 Billion polygons per second.

Ultimate scalability – Custom Supersystems scale as production systems to thousands of compute nodes and terabytes of main memory, which may be tightly coupled with LNXI FPGA or FPA for application specific acceleration. Leveraging robust parallel file systems, Custom Supersystems sustain Gigabytes of I/O throughput.

System-class Total Cost of Operation – LNXI applies a systems approach to custom supercomputing, delivering large-scale Custom Supersystems as fully integrated, validated and highly tuned systems. This extra effort on our part streamlines customer effort and costs for initial system deployment through ongoing system scale, operation and maintenance, dramatically improving Total Cost of Ownership (TCO) for these one-of-a-kind supercomputers.

Reliable Application Completion – Custom Supersystems feature a production-class environment that assures reliable application completion for extreme supercomputing applications. High-performance monitoring and intuitive system management tools assure that the system delivers reliable sustained performance and throughput throughout its lifetime.