Numbers are of prime importance in encryption.
Actually, prime numbers are important — very large ones are used to generate encryption keys. Numbers that can’t be divided by anything but themselves seem to appear at random intervals mathematically.
That’s why the University of Calgary
Centre for Information Security and Cryptography (CISaC) project to predict the fastest speed at which an algorithm can predict prime numbers is as cryptologically significant as it is mathematically harrowing.
“”We’re talking very large numbers,”” said Marc Wrubleski of the department of mathematics and statistics at the university. CISaC tested numbers up to 19 trillion trillion — that’s a 19 followed by 24 zeroes — on a Beowulf cluster of blade servers, taking 25 days of 186-CPU power. The number is equal to about three times the earth’s weight in pounds.
Other imposing calculations have included verifying new algorithms using slower but better-proven methods (two days, 256 CPUs) and improving algorithmic security by identifying large numbers with characteristics that make them particularly cryptologically significant (four weeks, 278 CPUs).
The fact that this high-intensity number-crunching is being performed on a cluster of blade servers shows how far in from the edge computing environment the blade has moved into the computing infrastructure, Wrubleski said.
Frank Morassutti, eServer xSeries specialist with IBM Canada Ltd., described the typical blade customer in Canada as investing in infrastructure and already in a 1U or 2U rack environment. Blades are being used in clustering, load-balancing and compute-intensive applications. Vertically, they’re in higher education, where lots of nodes are required for high-performance and clustering applications; in automotive, where there’s a substantial Citrix presence, which Morassutti said is ideally suited to a blade environment; and in life sciences, industrial applications and telcos.
The university has created a 139-node cluster for its number-crunching pursuits, each consisting of two 2.4 GHz Intel processors with 2GB of RAM. In the RFP process, the IBM system finished behind a white box offering in the price-performance scoring, but Wrubleski said the total cost of ownership story was even better. The blade system uses a lot of shared infrastructure — floppies, CD drives, power and cooling — that spell lower power consumption in the long run.
There’s also reduced complexity, Wrubleski said. The Calgary system uses fewer than 140 cables, compared to more than 600 in a 1U rack environment, and has a much smaller footprint.
The footprint was an important consideration for the City of Yellowknife when it poured federal Smart Community Project money into updating its infrastructure. Twenty-two server, some so old, they running on OS/2 and token ring, handled applications across departments. “”Server consolidation was a huge issue for us,”” said Ciaran Coates, systems analyst with the city. “”To have an individual server (for each application) is out of the question for us.””
But those 22 boxes couldn’t be shut down and moved while the new blade environment was installed, so the new system’s footprint had to be small.
The City also found the lower power consumption attractive — “”power being at a premium in the N.W.T.,”” as Coates noted — and its streamlined management, since the City runs a four-person IT shop. If senior staffers go on vacation, it helps “”if things are easy for the junior staff to take care of,”” he said.