Presentation on theme: "SUPERCOMPUTERS By: Cooper Couch. WHAT IS A SUPERCOMPUTER? In the most Basic sense a supercomputer is one, that is at the forefront of modern processing."— Presentation transcript:
WHAT IS A SUPERCOMPUTER? In the most Basic sense a supercomputer is one, that is at the forefront of modern processing capacity. A supercomputer can take on one of two configurations; the first is called “distributing computing,” the second is called a “cluster.” Distributed ComputingCluster This is when a large number of computers are distributed across a single network devote some, or all of their time to solving a common problem. A cluster is a when a large group of highly advanced computers are placed in close proximity to each other. An example is the internet being the network and all computer around the world being the distributed computers. An example would be the Blue Gene supercomputer at Argonne National Lab.
THE HISTORY OF SUPERCOMPUTERS Supercomputers have been around since the 1960’s. In fact most of the supercomputers in the world at that time were made by one man, Seymour Clay. The most influential supercomputer of the 1960’s was the CDC 6600. The CDC 6600 was 10x times faster than any other “supercomputer” at the time of its release. This computer helped to define and make the supercomputer as we know it today.
HISTORY CONTINUED Seymour Cray went on to build the Cray 1 and Cray 2 in the 1980’s; both are the fastest computers of their time. As the world entered the 1990’s supercomputers performance increased greatly. They went from only a few processors to thousands of processors seemingly overnight. Currently supercomputers are grouped in clusters with each individual cabinet of the cluster containing thousands of processors.
THE BIGGEST PROBLEM TO SUPERCOMPUTERS One problem has been with supercomputers since their beginnings back in the 1960’s, and it still haunts them now. This problem- heat. Heat is the biggest enemy to any computer, but especially supercomputers. Heat is such a problem because it can, and will, shorten the lifespan of the hardware. The reason this is important is because supercomputers rely on so many things that if one thing goes wrong with it, then the whole computer is out of commission.
HOW HEAT IS COMBATED Many different approaches over the years have been tried. From liquid cooled, to air cooled, even to hot water cooled. All of the different methods have pros as well as cons associated with them. Until the advent of modern day supercomputers which are placed in a cabinet style case with shelves holding thousands of processors, the common approach was liquid cooling with Flourient. This method is no longer feasible so in the last decade many new methods have arisen. IBM have employed a hot water cooling system in their Aqusar supercomputers. Other companies have tried using a combo’d version of liquid and air cooled, the most recent being one designed by the Leibert Company.
OPERATING SYSTEMS At the turn of the century operating systems for supercomputers made a big change. Early on the operating systems custom built for each supercomputer to maximize its own performance. This changed with the century, now supercomputers on run off generic software systems that are then configured to each supercomputer. The most popular of the systems among the top supercomputers in the Linux system. Originally this system, launched in 1991, was meant for windows x86 but over the years has been changed to be compatible with supercomputers.
HOW TO MEASURE PERFORMANCE ON A SUPERCOMPUTER? On modern day supercomputers performance is measured in FLOPS (Floating-point Operations Per Second). Currently Supercomputers are in the petaFLOPS range. This means that a Current supercomputer like the Titan in Oak Ridge, TN can reach 23x 10^15 FLOPS. Supers aim for maximum capability over capacity. This is because when using maximum capability it means the computer is using the maximum computing power at that time. Compared to capacity where its using cost- effective computer power.
PERFORMANCE CONTINUED Supercomputers across the world are benchmarked using what is called the Linpack Benchmark. This is a test which causes the computer to use its maximum capability to find out how fast a computer can solve a problem.
HOW SUPERCOMPUTERS ARE RANKED Since 1993 a ranking system derived using the Linpack Benchmark has been in effect ranking ever supercomputer in the world against each other to find the fast one in the world. This list is called the TOP500. Currently the number 1 spot on the list is held by the NUDT Tianhe-2 at 33.86 petaFLOPS, this is almost twice of the 2 nd place holder at 17.59 petaFLOPS.
APPLICATIONS Over the years supercomputer have had many different applications from weather forecasting to Molecular Dynamics Simulations. As the applications changed the performance, design, and usage of supercomputers changed with it. As the applications become more and more needing of greater computing power, the computing power of supercomputers will grow. It’s a natural progression in development that is thanks to humans need to know more.