ACADEMIA

Massively’s Biometrics Delivers High Performance, High-Volume Capabilities at a Fraction of the Cost of FPGA-Based Systems -- Massively Parallel Technologies, Inc. (Massively), a provider of on-demand high-performance computing (HPC), announces fingerprint match rates of 465,000 records per second using only its advanced, patented Howard software technology. Measuring true end-to-end match times (rather than just compute time, which does not reflect real-world performance requirements). Massively is expanding its already impressive performance by targeting 1-second match times running on commodity commercial off-the-shelf systems searching tens-of-millions to hundreds-of-millions of biometric records over the next few months. In comparison to costly and inflexible FPGA-based systems, Massively Biometrics delivers phenomenal match rates running on commercial off-the-shelf systems and is algorithm agnostic, which means customers can use any algorithm they wish, now or in the future. And Massively Biometrics delivers end-to-end speed and accuracy searching 100% of the target database and supporting 180-degrees of fingerprint rotation. Other biometric solutions report compute time only, limit input fingerprint rotation and/or search only a fraction of the database in order to report high match rates. But such “corner cutting” results misrepresent and disservice real-world deployments. “We believe that Massively has taken the first step towards demonstrating high performance, high accuracy biometric systems used for security and access verification can be delivered at a fraction of the cost of costly and inflexible FPGA-based systems” said Scott Smith, CEO of Massively. “The scalability and price/performance ratio we are demonstrating, combined with an open, non-proprietary, commodity COTS-based system architecture, is unprecedented.” Massively’s patented and patent-pending Howard parallel processing software technology makes such biometric matching performance and accuracy practical and affordable through both compute and cross-communication efficiencies that keep communication overhead growth rates to near zero as processing nodes are increased, even out to thousands of compute nodes. Initial scalability and performance measurements of Massively’s Biometric capabilities were run using only 80 single-CPU (Intel Pentium 4) Dell compute nodes searching a database of 1 million fingerprint records. Neurotechnologija’s VeriFinger SDK, a rapid fingerprint identification software development tool, was used as the matching algorithm and was provided by Fulcrum Strategic Partners, Inc., the sole US distributor of the full line of biometric software development tools from Neurotechnologija Ltd. Neurotechnologija’s Verifinger matching algorithm was deployed on the Howard platform in just a few days using Massively’s API-based parallelization model – all without any access to source code or source code modifications. “These repeatable, stable results on inexpensive commodity COTS equipment mean that the Biometric market can be expanded well beyond current users”, commented Kevin Howard, Massively’s CTO. “Massively and its partners are ready, willing and able to meet the most demanding Biometric application.” Supporting a variety of biometric matching and identification techniques including fingerprint identification and face identification, Massively’s Powered by Howard biometric solutions make it possible to rapidly authenticate or identify individuals at an extremely low cost per identification without sacrificing accuracy. As the total number of desired database records to search grows at an exponential rate, processing and compute scaling efficiencies are paramount to the success of a biometric solution. Massively’s Biometrics can be made available through its Internet accessible, on-demand supercomputing Biometric Virtual Power Centers or through on-premise installations.