Tuesday, August 31, 2010

Sequoia – the 20 petaflop supercomputer

A supercomputer is a computer that represents the frontline of current processing capacity, particularly speed of calculation. They are used for highly calculation - intensive tasks such as problems involving quantum physics, weather forecasting, climate research, molecular modeling and physical simulations such as simulation of airplanes in wind tunnels, simulation of the detonation of nuclear weapons, and research into nuclear fusion. The fastest supercomputer today is Cray’s Jaguar (1.7 petaflops), but in 2011 will be delivered 15 times faster IBM’s Sequoia (20 petaflops), which will provide more processing power than the entire list of Top500 supercomputers running today or the 2 million commercial laptops.

The US government has commissioned IBM to build a massive supercomputer. The Sequoia supercomputer was revealed in February 2009, and scheduled for delivery 2011 and operation in 2012. It will be able to perform at 20 petaflops, or 20.000 trillion floating point operations/sec.
Sequoia will represent a significant leap forward in compute power.  With top speeds of 20 petflops, Sequoia will be approximately 15 times faster than the Jaguar (1,7 petaflops), today’s most powerful supercomputer, and it will offer more processing power than the entire list of Top500 supercomputers running today. This will be achieved by future IBM BlueGene technology and by using 1.6 million IBM POWER processor cores (16 or 8-core Power Architecture processors built on a 45 nm fabrication process) and 1.6 petabytes of main memory, which will occupy 96 refrigerator sized server racks occupying just 3422 square feet.  

Sequoia will require 6 megawatts (6MW) of power for delivery of 3,050 calculations per watt of energy (or 3000 Mflops/watt), which is unprecedented level of energy efficiency compared to current supercomputer designs. The Sequoia system will deploy a state of the art switching infrastructure that will take advantage of advanced fiber optics at all levels. Operating system of this machine will be Linux.

Sequoia will primarily be used to ensure the safety and reliability of the US’s nuclear weapons stockpile. It will also be used for research into astronomy, energy, human genome science and climate change.

It will be built, tested and benchmarked in IBM’s Rochester, Minnesota plant, home of the Blue Gene class of supercomputers the company builds for ultra-scale computational applications. The hardware and software development will be provided by IBM engineers in Rochester and by researchers in IBM’s Yorktown Heights, N.Y. research lab, in partnership with the Lawrence Livermore National Lab and the Argonne National Lab.

Historical overview of supercomputer’s development

Early machines, produced in 1960s, were simply very fast scalar processors (CPUs - central processing unit) which processes one data item at a time.
In the 1970s, most supercomputers were dedicated to running a vector processor (also called array processor), wherein a single instruction operates simultaneously on multiple data items. The difference between scalar and vector processor is analogous to the difference between scalar and vector arithmetic.
The early and mid-1980s saw machines with a modest number of vector processors working in parallel to become the standard. Typical numbers of processors were 4 - 16.
In the later 1980s and 1990s, attention turned from vector processors to massive parallel processing systems with thousands of "ordinary" CPUs, some being commercial, “off the shelf” units and others being custom designs.
Today, parallel designs are based on "off the shelf" server-class microprocessors, such as the PowerPC, Opteron, or Intel Xeon, and coprocessors like NVIDIA Tesla GPGPUs, AMD GPUs, IBM Cell, FPGAs. Most modern supercomputers are now highly-tuned computer clusters using commodity processors combined with custom interconnects.

Jaguar, the current fastest supercomputer in the world with 1.7 petaflops, has 224,256 x86-based AMD Opteron processor cores.

related links:

No comments:

Post a Comment