Wednesday, August 7, 2013

Bigger, Faster, Cheaper

As databases get bigger, they take longer to search. For a while, and in many applications, nobody really cares. Does it really matter if a criminal database fingerprint search takes one second or 1.5 seconds? A city of 1.5 million people may arrest 40 people on a busy day. In cases like this, the limiting factor to how many times a process can be repeated isn't in the technology.

But if the world is headed the way many expect, biometric searches of large databases will be moving from applications where fractions of a second don't matter much, as in the case above, to something that looks a lot more like what banks or large web sites do: handle thousands of transactions per hour among thousands of users with both the quantity of transactions and users fluctuating wildly over the course of a day, and generally increasing over time. Now, how the search happens starts to matter a lot and technique starts to affect cost.

In the biometrics world, sensor and algorithm innovation get a lot of attention. Database architecture and search techniques don't. This press release from BIO-key is a refreshing change highlighting one technique programmers can use to cope with ever-larger biometric databases.

Accelerated Biometric Indexing Search: New Fingerprint Matcher Design Yields Higher Accuracy at Higher Speeds per Dollar Invested (Press Release via pr-inside.com)
"Using Commercial-Off-The-Shelf (COTS) products, BIO-key is expanding the way a biometric search can be performed which dramatically improves speed over conventional approaches. This revolution comes from the use of a highly parallel search architecture, allowing our solutions to perform faster and look deeper while improving speed and accuracy," stated Renat Zhdanov, PhD, Vice President, Chief Scientist, BIO-key International.

Initial tests of the new accelerated architecture show speed results of several millions matches per second, on a typical PC. This provides biometric search acceleration of several orders of magnitude on that PC alone. "These performance gains mean the required hardware and support costs for larger systems, or those heavily used in the Cloud from mobile devices or other sources, can now be greatly reduced, providing for thousands of times more throughput per dollar spent," stated Mira LaCous, Senior VP of Technology and Development, BIO-key International.