The Idaho National Laboratory (INL) had been designated as the lead laboratory for nuclear energy research in 2002. With this responsibility assigned, the INL needed to ensure that it had a highly secure, flexible and transparent collaborative computational capability that could be reached by its international partners - the Generation IV International Forum initiative (GIF), which was was started by the U.S. Department of Energy's Office of Nuclear Energy, Science and Technology to advance research in using nuclear energy.
INL CIO Dan Wickard explained, "...we were searching for an architecture to handle secure, serial or loosely parallel collaboration processing. It also needed to be affordable, easy to use and available now so we could be operating and productive within weeks."
As an experienced and satisfied user of the SolarisTM Operating System (Solaris OS), INL chose Sun to provide its grid computing solution for GIF and other key research computational requirements in an agreement lasting three years and valued at nearly US$2 million.
INL already had a significant Sun Microsystems base installed. It decided to add-on to its existing infrastructure and purchased 250 Sun FireTM V20z servers, 12 TBs of Sun StorEdgeTM 6320 arrays, Solaris 10 OS, Sun N1TM Grid Engine software, Sun StorEdgeTM QFS software, Sun StorEdgeTM SAM-FS software, StarOfficeTM and the Sun JavaTM Enterprise System, all of which were integrated and delivered by SunSM Customer Ready Systems (CRS). With the Sun CRS employed, INL knew it could rest assured that it would receive a solution that had been integrated, configured and tested at the factory. This would reduce the amount of time needed to deploy the solution and ensured configuration completeness.
According to INL's Chief IT Architect, Eric Greenwade, work was being accomplished using the grid solution within 28 hours of delivery. "The racks practically came off the truck running; we were able to have the entire eight rack cluster up and submitting jobs within two days," he explained. "All we had to do was to hook it up to power and our network, boot the systems up, and we were ready to go."
The end-to-end high performance technical computing solution gives INL the ability to complete two trillion (2 Teraflops) floating-point operations in a one-second.
"Sun's grid computing cluster is a major component in advancing the large scale computing capability our researchers need for the design of the Department of Energy's Generation IV nuclear reactors," Greenwade said. "This cluster solution is a truly significant step forward in the demanding collaborative environment required among the 11 members contributing to the Generation IV design efforts."
GIF members include Argentina, Brazil, Canada, Euratom (a European consortium), France, Japan, Republic of South Africa, Republic of Korea, Switzerland, United Kingdom and the United States.
The 250 Sun Fire V20z servers, powered by dual AMD Opteron processors offered INL 64-bit performance, large memory, compute-cluster readiness and native high performance 32-bit x86 capability. The Sun Fire V20z server is well suited for INL's computational-intensive applications.
"The Sun Opteron processors run anywhere from three to ten times faster than what we had before," Greenwade explained. "That translates to substantial growth in work productivity. For instance, it took a nuclear engineer using a desktop solution four weeks to do one calculation on a single fuel pellet (the core part of a nuclear reactor). Now, we can work simultaneously on models, which are several thousand times larger than our previous ability. The end result is we can do problems that are up to 10 to 100 times bigger and get them done in as little as 1/10th the time and our confidence in results also has vastly improved."
Another area of vast improvement with this Sun solution was the high performance data management provided by two Sun FireTM 4800 servers that use fibre channel connections to ferry data back and forth from the 12 TBs of Sun StorEdge 6320 and existing 5 TB Sun StorEdgeTM 3510 FC arrays. Running Sun's StorEdge QFS software that lowers total cost of ownership by increasing the system administrator's span of control over a continuously growing data environment, this Sun solution comprised of a high performance, shared file system with fully integrated volume management, direct I/O and other features gives INL users the ability to distribute information quickly and easily while managing data from a central location.
Security also was high on the list of priorities for INL and its GIF partners. Sun's Grid Computing Solution provided the security levels required by laboratory through the Solaris 10 OS with SolarisTM Containers. "The Solaris Containers help us subset the data using zones for a 'project by project' approach or on a 'collaboration by collaboration' basis," underscored Wickard. "This will be invaluable to us in managing intellectual property."
The Sun N1 Grid Engine software is helping the laboratory continue to improve on its utilization rates by automating job allocation, finding idle compute resources and harnessing them productively. It delivers to the laboratory several times the usable power out of systems on the network.
"We decide what projects get the compute power and when, determined by our mission priorities," Greenwade said. "N1 Grid Engine helps us allocate resources much more effectively and efficiently, resulting in higher utilization rates."
For the future, INL would like to access the grid computing environment and plans to investigate the use of the Java Enterprise System to develop a customized grid computing portal for single sign on (SSO) by the user. Sun JavaTM System Portal Server, Sun JavaTM System Identity Server, Sun JavaTM System Directory Server and Sun JavaTM System Web Server may contribute to this secure grid computing portal deployment.
"Identity management is a key element for us and we need a way to carry the users' access along with their identity, so it can be managed centrally, and updated quickly," explained Greenwade. "The Sun solution offers a strength through the combination of security and flexibility of operation, so that we can provide a greater collaboration environment for the Generation IV nuclear reactor design project and other energy research."
Read More ...