Supercomputing tentacles of the ‘Octopus’ embrace the Technion

As a new member of PRACE, The Partnership for Advanced Computing in Europe, Israel has had to step up its computing capacity in order to participate in the group’s cutting-edge projects. As of this week, the Technion is ready to compute with the best of them, with the installation of an SGI (Silicon Graphics International) cluster supercomputer, affectionately known as “Tamnun” (Octopus).

Supercomputers differ from plain old computers in processor speed, capacity, storage, memory size, and the ability to handle very advanced calculations and computations. The processors (such as Intel Xeon-E5 processors) used in supercomputers are usually faster and have more capabilities — e.g., the ability to do multitasking — than standard home computers.

But what most differentiates supercomputers from their ordinary brothers is their ability to pool the resources of many processors, working together in order to make a single computation. For example, the fastest supercomputer today, the Sequoia, built by IBM for the US National Nuclear Security Administration, consists of 98,304 computing nodes containing 1,572,864 processor cores, and covers about 3,000 square feet of space.

The Technion’s new addition isn’t quite the Sequoia, but it is the most powerful computing cluster ever deployed for civilian use in Israel, said TNN Telecom, SGI’s representative in Israel, which deployed the system for the Technion. The Tamnun is an SGI Infiniband Cluster, consisting of 1260 processor cores, with 96 GB of RAM memory per node. The processors are “united” by SGI’s InfiniBand communication protocol. The system also features advanced graphic processing units, and can store 60 terabytes of data, with optional additional storage added as needed.

The primary users of the system will be faculty and staff of the Technion’s Minerva Optimization Center, which researches applications of optimization models and techniques in engineering, medicine, industry, etc., and the Russell Berrie Nanotechnology Institute, with members of other departments using the system less frequently.

Tamnun will enable Technion researchers to, among other things, participate in advanced projects at PRACE, where the projects are measured in terms of “core hours” — the amount of time spent on processing in the cluster — with each project requiring 100,000 hours and up. Obviously, the bigger the cluster, the better.

Speaking at the installation ceremony this week, Dr. Daniel Rittel, a senior vice president at the Technion, said that the installation of the system “puts us at a par with the most advanced academic institutions in the world.”