MCNC (www.mcnc.org) is a nonprofit organization located in Research Triangle Park, North Carolina. It was founded by the state General Assembly in 1980 to drive technology-based economic development in North Carolina by partnering with universities, businesses, and government.

MCNC promotes economic development in North Carolina through three lines of business - Research and Development, Venture Funding, and Grid Computing and Networking Services. Their grid initiatives are offered on the Linux platform and offer a unique opportunity for many research organizations. Specifically, MCNC allows research organizations whose budgets cannot justify a supercomputing infrastructure to still have access to these important tools that allow researchers to process large amounts of data and calculations.

Anatomy of the MCNC Grid
MCNC, through their Grid Computing Services Division, offers enterprise grid computing services for research, education, and commercial businesses. These solutions are offered via an "on-demand" model that allows users to have access to these powerful computing systems over the network. At the core of the grid are high-performance computing systems running Red Hat Linux (www.redhat.com). The systems that power the grid look more like video game cabinets in a state-of-the-art data center. The first is a purple SGI system that is a Symmetric Multi-Processor (SMP) SGI server with 32 Intel 1.3 GHz, 64-bit CPUs. The other, physically larger, system is a 64-node Massively Parallel Processor (MPP) IBM eServer cluster with a total of 128 Intel 2.8-GHz, 32-bit CPUs.

Building on a testbed grid that extends to North Carolina State University, the University of North Carolina at Chapel Hill, and Duke University, these systems will eventually be part of a statewide grid infrastructure that is being deployed across the MCNC-operated North Carolina Research and Education Network (NCREN). This statewide grid will allow even the non-research-intensive schools throughout the state to have access to computing resources that would be cost prohibitive to implement on their own campuses.

Why Linux?
Linux is the primary development platform for grid middleware, and it is becoming the operating system of choice for supercomputing applications according to Chuck Kesler, MCNC's program manager for Grid Deployment and Data Center Services. Kesler who has worked with both Linux and commercial Unix sees Linux as a viable platform on a par with systems that cost significantly more. He notes that they do still pay for services related to Red Hat Linux, but that Red Hat combined with Intel hardware offers the best price and performance of any solution they've found. Additionally, Kesler says that a deciding factor is that researchers are also increasingly using Linux in their own labs, which in turn is driving software developers to add Linux support for their scientific applications.

How Is Linux Being Used?
Currently, MCNC's Linux-based enterprise grid resources are being used primarily for molecular and atmospheric modeling applications that traditionally run in supercomputing environments. In addition, MCNC is working with its university and vendor partners to develop true grid applications on Linux. With a focus on bioinformatics research, two of their first successful grid projects are located in the Research Triangle area of North Carolina. NC State uses BioPerl (www.bioperl.org), a set of tools used in bioinformatics research powered by the MCNC Bioinformatics grid. At the UNC, researchers are using IBM's WebSphere (www.ibm.com/websphere) to facilitate data flow for molecular modeling used in drug discovery. These initial products are demonstrating to other researchers the cost-effective Linux solution that may help them with their next big technology breakthrough.

A Commercial and Open Source Mix
Based on an architecture developed in collaboration between MCNC, NC State, UNC, and Duke, MCNC has implemented a system that melds open source and commercial grid middleware packages. Kesler says that the architecture team looked at the Open Source Globus Project (www.globus.org), which provides a widely accepted toolkit for building grids, but found that it did not deliver a complete solution for data management. They turned to Avaki (www.avaki.com) to fill this gap. By creating a federation of data resources that are contributed to the grid by multiple parties, Avaki allowed MCNC to build a global filesystem that spans organizational boundaries. This type of approach is contrary to a traditional data structure where accessing islands of data across organizations requires significant effort.

Linux and Grid Computing Impact Business
MCNC is tasked with economic development as part of their charter. David Rizzo, CEO and president of MCNC, says that MCNC is providing grid computing solutions to businesses in the midrange manufacturing sector, allowing them to participate in supply-chain activities that without this resource they could not participate in. Rizzo notes that before they started to use Linux as their platform of choice they used proprietary operating systems and hardware.

The impact of their initiative is far reaching especially for the state of North Carolina's economy, which includes a mid-tier manufacturing base. A study released by the Rural Internet Access Authority estimates grid computing could positively impact the state economy by as a much as $10 billion by 2010 through job creation and increased commerce. The engine that drives this grid computing will most likely rest on Linux.

The Future of Grid Computing
Grid computing driven by low-cost Linux operating systems is an important part of the future of IT data centers. Not only are the systems scalable, they leverage low-cost Intel hardware that can easily scale to offer massive supercomputers at a fraction of the cost of proprietary hardware and operating systems.

Financial services, life sciences, government, and aerospace industries will all benefit from successful implementation of these massive computing powerhouses. Pilot programs like the MCNC initiative will also help the proliferation of grid computing as they pioneer the development of programming tools, applications, and management systems to allow the research organizations in their network to utilize these powerful tools.