4-page Case Study - Posted 3/13/2007
Views: 1346
Rate This Evidence:

Boeing

Boeing Tests High-Performance Computing Cluster, Improves Processing Time

The Distributed Software and Systems Integrations (DSSI) group in the Phantom Works advanced research and development organization at Boeing, a global aerospace company, wanted faster processing of finite element analysis projects and a platform that could handle a spectrum of applications. Engineers processed jobs on local workstations and sought a solution that would speed up service delivery to internal customers. Boeing required a platform that would be easier to use and could accommodate a wide variety of Windows® -based applications. The company installed Windows Compute Cluster Server 2003 on an HP Cluster Platform 4000 in a lab test-bed environment to evaluate application compatibility and performance.

Situation

Headquartered in Chicago, Boeing is a global aerospace company and manufacturer of commercial and military aircraft. The company employs more than 155,000 people in 67 countries. Boeing has a long tradition of innovation, and its research and development (R&D) groups play a central role in helping the company keep its competitive edge.

*
* It is important that our IT environment is easy to use and support. Windows Compute Cluster Server on the HP Cluster Platform has potential for improving our performance and manageability.  *
Dr. J.S. Hurley
Senior Manager, Distributed Software and Systems Integration
Boeing
*
Boeing R&D teams, such as the Intelligent Graphics group in Phantom Works’ Mathematics & Computing Technology organization, process jobs for internal design teams who then use the analysis in commercial production or in other development activities. Engineers in these R&D groups work with an assortment of intelligent graphics analysis and recognition tools to create high-vector illustrations that respond to user actions such as clicking hyperlinks and that can be used as part of documentation sets and manuals. In addition, many research engineers use finite element analysis applications such as MSC Software’s Nastran 2005R3b and Abaqus 6.6-4 Standard version for performing non-linear static analysis used to verify the structural integrity of aircraft.

Intelligent graphics recognition and analyses can be extremely daunting given that documentation for each airplane typically involves more than 20,000 graphic depictions and nearly a million pages of cross-referencing information. Using complex graphics recognition software, every graphic must be converted into a searchable, navigable electronic format with embedded technical information. Engineers were running these computationally intensive applications on stand-alone workstations, and, as a result, files could take days to process. Such challenges were equally familiar to groups that work with finite element analysis software. In addition, engineers invested considerable time monitoring applications, and their customers faced potential delays to production as they waited for results.

Boeing Phantom Works’ Distributed Software and Systems Integrations (DSSI) group helps R&D groups work more efficiently through the use of new and emerging computing technologies. To improve productivity for Boeing’s R&D teams, The DSSI group began looking for a solution that could speed processing of engineering applications. Engineers hoped that faster processing speeds would give them better control over projects. “We wanted to view dynamic changes in real time,” says John F. Bremer, Advanced Computing Technologist in the Intelligent Graphics Group. “We thought that if we could reduce processing time to a few minutes, then we should be able to observe any changes we make to the program or data files immediately. At the moment, we have to wait for hours or even days to get the results back.”

Jon B. Franklin, Boeing HDC Stress Analyst/High-Performance Computing Administrator, had similar concerns. Franklin runs Nastran and Abaqus applications in a workstation environment. “We do finite element analysis, and, as an end user, I had limited resources,” says Franklin. “I had a limited number of machines, but I thought that if I could convert them into a high-performance computing cluster, my results would be obtained faster and I could save time and money.”

Improving performance was not the only priority of the R&D groups. Above all, the engineers wanted a platform that would accommodate a variety of engineering applications, yet also enable other activities. “We needed our workstations to function as both part of a cluster and as a stand-alone compute environment where we could do other tasks,” says Franklin. “We didn’t want to have computing devices segregated and not be able to log on and work with standard Windows®-based applications.”

The DSSI group realized that a high-performance computing (HPC) solution that could handle a wide variety of engineering applications and deliver improved performance was needed. It also sought to standardize the various R&D groups on a platform that would be easy to use and work well with the existing IT environment.

Solution

As the DSSI group began exploring various HPC options in early 2006, it found widespread interest at Boeing in a Microsoft® approach to high-performance computing. Many groups were already using existing Windows operating system tools and thought that porting the applications to other platforms would be time consuming and not an optimal approach. Instead, Boeing engineers preferred the idea of an HPC resource with a familiar, accessible user interface and management tools. Bremer says, “Speed is important, yet I would also say a user-friendly interface is necessary. If it’s not something that most engineers can work with comfortably, then it won’t be used.”

As a result of this input and its own research into alternative solutions, the DSSI group decided to install Microsoft’s Windows Compute Cluster Server 2003 in a test-bed environment. Windows Compute Cluster Server is built on the Microsoft Windows Server® 2003 x64 Standard Edition operating system, and is designed to support high-performance technical and scientific applications that take advantage of parallel processing.

*
* Because we develop intelligent graphics software in a Windows environment, it makes sense to work with an HPC cluster that supports that framework.  *
John F. Bremer
Advanced Computing Technologist, Intelligent Graphics Group
Boeing
*
Boeing knew that installing Windows Compute Cluster Server would give engineers the benefit of extending the familiar Windows desktop environment into the world of high-performance computing. Bremer says, “Because we develop intelligent graphics software in a Windows environment, it makes sense to work with an HPC cluster that supports that framework. Otherwise, we incur a lot of expense and effort in porting applications to a different platform.”
 
In August 2006, the group installed the Windows Compute Cluster Server on HP Cluster Platform 4000 hardware. The HP platform was chosen because HP’s Unified Cluster Portfolio offers a comprehensive, modular approach to building computing clusters. The modular approach gives customers flexibility by reducing complicated cluster hardware design, configuration, and deployment issues. “We found the HP Cluster Platform very useful because it provided defined cluster configurations with the flexibility to choose the types of servers, the number of servers, and the interconnects,” says Dr. J.S. Hurley, Senior Manager, Distributed Software and Systems Integration.

“In addition, the cluster is built at HP’s manufacturing facility,” continues Hurley. “The fact that we have the flexibility to preload Windows Compute Cluster Server on the hardware at the HP facility provided significant advantages in deployment.”

Implementation of the new cluster was further simplified by using Windows-based application deployment tools. The DSSI group used the installation wizards of Windows Compute Cluster Server to automate deployment, and was able to install and integrate the new cluster with the corporate Active Directory® service quickly. “Setup is pretty simple,” says David Shaw, Advanced Computing Technologist for Distributed Software and Systems Integration. “As long as you have the architecture defined, all you have to do is integrate it with Active Directory and it’s ready to go.” Gregg Sherles and Ron Steinke of the Boeing Information Technology organization estimate that once the hardware is properly configured, deployment of the entire cluster takes about two hours to complete using images created in the Windows Preinstallation Environment. The group chose to work in the Windows Preinstallation Environment to reduce downtime and allow engineers to begin testing applications quicker.

The DSSI group began testing critical software immediately, starting with Nastran 2005R3b. The current SMP (shared-memory parallel) version fared well on the 64-bit operating system and HP Cluster Platform. Hurley, who led the Case Study testing, noted, “We saw good improvement on the Nastran jobs, and we’re preparing to test the early release of Nastran DMP (distributed-memory parallel).” The most recent release of Nastran DMP software will be optimized for distributed processing on a Windows platform.

Engineers also tested intelligent graphics diagrams in a simulated production environment to address scalability requirements. John Bremer explains the reason for using a generic prototype: “In our actual workflow, we have anywhere from 4,000 to 8,000 different diagrams that we process, but because of the sensitivity of those diagrams, we couldn’t use them for testing. We had to use a scrubbed, or modified, file that we replicated up to 10,000 times to get an idea of the scaling.”

The group uses the integrated Job Scheduler of Windows Compute Cluster Server to queue jobs and distribute the workload across the cluster. Before, engineers had to enter each job separately into individual workstations. The engineers also appreciate that the Job Scheduler interface is easy to use, with menus and navigation features already familiar to people accustomed to working in a Windows environment.

In addition, the Intelligent Graphics group is exploring the possibility of developing customized applications for the new cluster. Many Boeing engineers already work with the Microsoft Visual Studio® 2005 development system to design applications used in production and could compile code for the HPC cluster as well. Bremer says, “Even now we have applications that are pushing the memory and computing limits of the workstations. There could be considerable benefit to developing some of those on the cluster.”

Preliminary tests have shown a significant reduction in processing time for some complex analyses, and notable improvement in processing speed for intelligent graphics recognition. The tests were conducted with a limited set of HPC analysis tools. Additional testing is required to determine if this configuration can satisfy other HPC requirements. In addition, a comparative analysis will need to be performed between this configuration and UNIX/Linux solutions. The DSSI group will continue to test the cluster, and will make the results of this research available to other groups at Boeing. These groups in turn will use the information gathered by the group to facilitate installation of their own Windows-based HPC clusters.

Benefits

By installing Windows Compute Cluster Server, Boeing engineers predict significant improvement in processing time for projects. The workflows of developers and their internal customers will be expedited as a result. In addition, the entire enterprise will benefit from having an easy-to-use HPC resource that integrates easily into its existing IT environment.

Reduces Processing Time

*
* We needed our workstations to function as both part of a cluster and as a stand-alone compute environment where we could do other tasks.  *
Jon B. Franklin
Boeing HDC Stress Analyst/High-Performance Computing Administrator
Boeing
*
Boeing engineers have already observed considerable gains in performance and speed, and expect to see even further improvement as more engineering software is optimized for distributed processing on a Windows platform. The DSSI group significantly reduced processing time for the Nastran jobs on the SMP version of the application and is very optimistic that even better performance will be possible with the DMP version.

Although the HPC cluster is still operating in a test environment, the engineers see a clear benefit in moving to a Windows-based HPC cluster. Bremer says, “This is the first Windows based technology that our group had access to at Boeing to do HPC processing. We saw orders of magnitude of improvement in speed, so it’s very exciting from that standpoint.”

Improves Workflow, Productivity

Faster processing speeds will improve the workflow and productivity for more than just the R&D groups at Boeing. Because jobs complete quicker when using the cluster, research engineers can take on more projects and deliver faster results to customers. “The workflow of a lot of people will improve dramatically if it takes a couple of minutes rather than several hours to process a set of graphics,” Bremer explains. “Developers will spend less time monitoring computers, and our internal customers can get their jobs into production more quickly.”

Offers a standardized platform that’s easier to use, manage

The DSSI group knew that the only viable solution for the needs of the various user communities would be an accessible, easy-to-use HPC cluster. Says Hurley, “It is important that our IT environment is easy to use and support. Windows Compute Cluster Server on the HP Cluster Platform has potential for improving our performance and manageability.”

For More Information

For more information about Microsoft products and services, call the Microsoft Sales Information Center at (800) 426-9400. In Canada, call the Microsoft Canada Information Centre at (877) 568-2495. Customers who are deaf or hard-of-hearing can reach Microsoft text telephone (TTY/TDD) services at (800) 892-5234 in the United States or (905) 568-9641 in Canada. Outside the 50 United States and Canada, please contact your local Microsoft subsidiary. To access information using the World Wide Web, go to:
www.microsoft.com

For more information about Boeing products and services, call (312) 544-2000 or visit the Web site at:
www.boeing.com

Microsoft Server Product Portfolio

For more information about the Microsoft server product portfolio, go to:
www.microsoft.com/servers/default.mspx

This case study is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY.
Document published March 2007
Solution Overview



Organization Size: 155000 employees

Organization Profile

Boeing is a global aerospace company with more than 155,000 employees and customers in 145 countries. Head-quartered in Chicago, the company’s annual revenues exceed U.S.$54.8 billion.


Business Situation

Engineers at Boeing work with complex files that were taking days to process. They looked for a high-performance computing solution that would be easier to use and that would improve processing speed.


Solution

To improve productivity and process jobs quicker, Boeing engineers installed Windows® Compute Cluster Server 2003 on an HP Cluster Platform 4000 for use with their finite element analysis and intelligent graphics applications.


Benefits
  • Reduces processing time
  • Improves workflow, productivity
  • Offers a standardized platform that’s easier to use, manage

Hardware

HP Cluster Platform 4000


Software and Services
  • Microsoft Visual Studio 2005
  • Microsoft Windows Compute Cluster Server 2003
  • Microsoft Active Directory Domain Services

Vertical Industries
  • Aerospace Industry
  • Defense Industry

Country/Region
United States