It was a spectacular feat, especially for those who swear by the Mac.

Today, many in the academic and other communities have expressed an interest in creating their own G5 superclusters. And now it is official: the Big Mac, with its 10 plus teraflop speed, had vaulted into the rank of third fastest supercomputer in the world, according to an announcement November 17 at the Supercomputing Conference 2003 in Phoenix.

But unknown to most in the news media was the fact that one of the reasons why Varadarajan could tweak that cluster of Mac computers had to do with software developed by another Indian American computer scientist.

Varadarajan ported software developed for the Unix system by Prof. Dhabaleswar Panda at Ohio State University and optimized the cache memory management of the G5. It turns out that the software developed by Panda, called MVAPICH, works by connecting traditional supercomputing software with innovative networking technology that speeds data flow. The Panda team included graduate student Sushmita P. Kini.

MVAPICH bridges the gap between the traditional message passing interface (MPI) -- the software that manages communication between nodes on a supercomputer -- and the InfiniBand technology. InfiniBand, short for “infinite bandwidth,” is a new networking architecture standard that was developed by an industry consortium to support high performance computing systems, including supercomputers.

Until Panda and his team at the Ohio Supercomputer Center, developed MVAPICH in 2002, InfiniBand and MPI were hopelessly incompatible, Panda said. The name MVAPICH is short for “MPI for InfiniBand on VAPI Layer.” VAPI refers to the VAPI software interface developed by Mellanox. MVAPICH is pronounced like “em-vah-peach.”

For the standard NAS Benchmarks, a commonly used benchmark suite in the high performance computing community, the MVAPICH software with InfiniBand reduces parallel execution times of some applications up to 33% when compared to other proprietary networking technologies. This groundwork opens the door for market entry of InfiniBand hardware into high performance computing.

While supercomputers were once built only as large-scale mainframe structures that were extremely expensive -- some costing tens to hundreds of millions of dollars -- a different kind of supercomputer based on clusters of many desktop-style computers has become more common in recent years, explained Panda.

Several system components contribute to supercomputing computation speed, such as the software used, the networking platform in place and the computer hardware that supports the network. Researchers continuously work to improve the performance of these components with the overall goal of reducing computation time by developing new supercomputing technologies. However, end users cannot fully reap the benefits of these new technologies until solutions are created that allow these technologies to work together.

Until MVAPICH came along in 2002, the complex mathematical problems that employ supercomputers, such as atomic modeling, decoding the human genome or molecular drug designing, posed a substantial problem for cluster computing. That’s because individual computers, called nodes, must compute in a parallel manner while sending much information back and forth to each other. Panda’s software program helps narrow the gap.

Parallel computing, the use of two or more computers to solve a single problem, enables researchers to break large problems down to smaller pieces that can be solved in parallel -- a big time saver. The computer network receives its directions on how to break the problem into smaller pieces from parallel computing software programs written by the supercomputer user. MPI, or the message passing interface standard, has become the de-facto choice for writing and developing parallel programs for supercomputing applications.

MPI has been developed over the years to write parallel programs on distributed-memory parallel systems that are portable across different types of computer systems, thus providing supercomputer users with more options. Prior to this, supercomputer users had to write separate programs for applications for specific supercomputing systems

As beneficial as MPI has proven to be for parallel computing applications, it cannot deliver performance unless protocols are known for each type of networking technology that can be employed by a supercomputer network.

“Even though basic MPI can work on naïve networking protocols like TCP/IP, it does not deliver performance unless the protocol stack of MPI is redesigned and implemented over the lower-level communication interfaces of a network while trying to take advantage of its features,” said Panda. “Thus, as new networking technologies evolve, it becomes a challenge to redesign and implement the MPI protocol stack to reap the benefits of the new networking technologies, such as InfiniBand."

InfiniBand, short for infinite bandwidth, is a new generation networking architecture standard that was developed by an industry consortium to provide high performance and scalable networking solutions for next generation high-end computing systems, including supercomputers.

Noticing the technology gap, Panda and colleagues conceptualized a design and implementation of a new interface to MPI that would allow the benefits of MPI and InfiniBand to be simultaneously exploited. If successful, it would also provide an opportunity for supercomputer users to evaluate and take advantage of the new InfiniBand networking architecture.

“At some point, adding nodes to a cluster doesn’t make the calculations go any faster, because it introduces communication and synchronization overheads, and researchers have to rely on software to manage communication between nodes effectively,” Panda said. “MVAPICH takes that software a step further by connecting it with the emerging InfiniBand network technology.”

A collaboration with computer chip maker Intel and leading InfiniBand developer Mellanox Technologies, Inc, of Santa Clara, California, is opening the Ohio State software to further applications in research and business. These companies have used MVAPICH to enable calculations on an off-the-shelf supercomputer that is capable of performing teraflop-level computing, or trillions of calculations per second. Intel calls-the system TOTS, for “TeraFlop-Off-the-Shelf,” and it will debut in the exhibition hall at the supercomputing conference.

Panda believes that the development of TOTS is leading to a new era of systems where research labs and commercial companies with smaller budgets can benefit from supercomputing technology. MVAPICH helps to make that happen, he said.

Since 2002, more than 65 organizations world-wide have downloaded the open source MVAPICH code to develop applications. One of the first was the weapons lab, Sandia National Laboratory, which recently used MVAPICH to power a large-scale (128-node) supercomputer. A similar project at Los Alamos National Laboratory involves a 256-node supercomputer. They plan to use Panda’s MVAPICH software for their experimentation and evaluation.

“These projects at national labs are important, because they show that our software can scale up from small applications to large,” Panda said.

Panda is continuing his research to provide other critical interfaces in MVAPICH to take advantage of many novel features of InfiniBand. These interfaces include support for collective communication and quality of service. Collaborations have already begun to design next generation MPI, called MPI-2, on InfiniBand as well as other programming interfaces. Panda is also designing other kinds of systems for high-end computing, such as distributed shared memory, parallel file systems, web servers, and internet data centers, to exploit the benefits of InfiniBand.

It is hoped that the current MVAPICH work and its follow-up research directions will not only impact the computation speed of supercomputer powered research, but eventually enable society to solve the complex problems that are requiring increased computation speed.

Dhabaleswar K. Panda obtained his Ph.D. in computer engineering from the University of Southern California in 1991. He earned his B.Tech (1984) from IIT Kanpur and his M.S. (1986) from the Indian Institute of Science, Bangalore, working briefly at ISRO and WIPRO. His research interests include parallel computer architecture, user-level communication protocols, interprocessor communication and synchronization, network-based computing, and high-performance computing.