Simpson Strong-Tie, Tabor Research's current featured user site, is a leading designer and manufacturer of construction products and a relatively new user of High Productivity Computing (HPC) products. Though new to HPC, Simpson has managed to leverage a small Linux Networx cluster (5 nodes, 56 GB of memory) in ways that have allowed the company to reduce its products' time to market and increase its product diversity. We believe that Simpson's experience is important because it defines several major characteristics of today's market that tend to be obscured by the more dazzling high-end applications and the race for the petaflop.

High Technology Is Everywhere

Every home owner can identify with the importance of keeping a roof attached to a house during a hurricane, a house attached to its foundation during an earthquake, or simply not having to worry about a structural repair because there was no product available to effectively address the problem. Yet the number of design factors and trade-offs involved in developing even the most commonplace connector can be mind boggling. Factors, range from the properties of the nails that might be used, the characteristics of wood or concrete, the properties of the connector materials under load, or the ease of installation and manufacture. From a design and testing point of view Simpson Strong-Tie is a high-technology company, facing standard business requirements to bring new and more capable products to market as quickly and cost effectively as possible. Thus, it is not surprising to see similar companies look to computational tools as a logical step in meeting these requirements, and improving their competitive positions.

Little Big Computing

Back of the envelope calculations and a check of the TOP500 website indicate that the Simpson Strong-Tie's machine has roughly equivalent in peak performance to top line machines of the early 1990's. Since that time, an organization's access to the highest levels of computing power has decreased -- top line HPC machines cost in the tens of millions of dollars -- and are only available to the largest institutions. That being said, systems from the first half of the last decade were used successfully in aircraft and automobile design and products based on those designs are still in the air and on the road. We continue to trust our lives to those designs to this day. Now this level of computing power is available to a broad range of small and medium businesses, departments, and individuals.

Entry-Level Potential

Entry-level HPC computing is the most dynamic and least understood segment of the market. It has been a major growth driver for the market for several years, and promises continued expansion as the number of first time industrial users grows. However, market requirements are confounded by the number and variety of organizations that are potential candidates for HPC solutions. Simpson Strong-Tie's experience does reaffirm some basic principles in selling to this market:

• Solutions sale -- The Linux Networx cluster in use is a complete self contained HPC solution including: service and compute nodes, direct attached short term storage, networked long term storage, applications software (ABAQUS), system management software (LSF from Platform Computing), and network communications to users. This is all in one rack, has room to grow, and is managed by the primary user.
  
  
• Applications support -- Knowledge of and ability to provide basic integration support for the primary application code is often a prerequisite for competing in these markets.
  
  
• System support -- A system architecture that allows users to perform basic support operations (e.g. swap out nodes), combined with strong backing of the product, including: fast turn around of failed nodes and easy management of the overall maintenance process are important factors in providing affordable efficient maintenance, while minimizing users' time spent in system manager mode (as opposed to design engineer mode).