Collaboration is one issue where end-users have been pushing hard to make progress. Intel, IBM, Nokia and other firms with large numbers of designers want their design groups to use a collaborative environment so designers in different locations can work as a single team. Auto firms and aerospace companies have also been facing the issue of collaboration, largely because they need to reduce the time-to-market for new products (note the disastrous consequences in the Airbus case!).
Recent and forthcoming meetings of European Commission-funded projects (e.g., SIMDAT, VIVACE) and of service-oriented architecture (SOA) product firms (e.g., the MSC Software conferences on virtual product development) indicate that auto and aerospace firms have made real progress. Using commercial SOA products or tools and grids, they have created “virtual cars” and “virtual planes,” 3-D files with data that support collaboration between engineering design groups. The virtual cars test designs and manage the inclusion of crash, computational fluid dynamics (CFD), structural testing, and noise-vibration and harshness modeling in the overall model of a car.
The advantages of using “virtual cars” are several. First, firms can go through many more static and dynamic simulations using well-known tools such as LS Dyna, NASTRAN, etc., for crash and CFD analysis. This provides robust results and shortens time-to-market. Second, testing and design costs are reduced because simulations replace much of the work that used to be done in these groups. This not only reduces costs, but also cuts down what had been a major area in product development. Third, auto firms are looking for more flexibility in design. They need to be able to respond quickly to new regulations, such as German requirements that the front ends of cars create fewer significant injuries to pedestrians. Virtual cars have let manufacturers change a model design quickly. Fourth, virtual cars have let companies such as Audi greatly increase the number of models they offer. They can design different versions that build on the same frame and often retain other parts of an established design, such as the windows, doors or trunk. Audi now markets about 40 models compared to about six a decade ago.
To create virtual cars, auto firms have enlarged their grids to more than 1,000 CPUs. They have employed traditional applications as part of their SOAs, and employ many simulation and CAE applications on SOAs. Having excellent access to needed data, as well as to compute power, is a sine qua non for virtual cars.
At the present time, virtual cars are developed at a single simulation and design facility. Because virtual car files are often at the 10 terabit level, firms have begun to expand the capacities of their campus WANs. In the future, they plan to link distant design centers together. Design engineers at Audi in Ingolstadt, Germany, will work with their counterparts at SEAT in Martorell, Spain, on the same “virtual car.” Engineering groups, such as Berco, that do undercarriage parts subcontracting may also be linked to the Audi network to collaborate in engineering design. This sharing will require high-speed networks with capacities well beyond the 10-20 Mbps now used to communicate with partners. Dark fiber is one way to put these connections in place.
However, not all auto firms have been moving in this direction. Some have had difficulty using virtual cars. In some cases, this is because there is a true divide among big auto firms. Some have focused on lean manufacturing, almost to the exclusion of what might be called “lean” engineering (i.e., flexible and visually beautiful design that also greatly shortens product development time and reduces development costs). In some auto firms that don’t use the tools SOA software firms have chosen to support “virtual cars,” they have found it a very great challenge to use this innovation. It remains to be seen whether they will continue to resist changing their design operations or if “lean manufacturing” [creating the greatest efficiencies in manufacturing] will trump “lean design.” Initial interviews suggest that it might not be possible to compete in global auto markets unless you combine “lean manufacturing” with “lean design” by using virtual cars.
Virtual cars could well be the tool that expands the use of grids. If auto firms can use virtual cars, and aerospace firms begin to rely upon virtual planes, the way they do collaboration could suggest solutions for the EDA firms such as Intel and IBM. It might even provide a vehicle for pharmaceutical firms doing drug design to come up with 3-D models of drugs that can be worked on by groups throughout the world. Thus, virtual cars could push grids to a new frontier, the “collaboration grid.” With collaboration grids, different simulation and design groups could operate as a single team in spite of the fact that they are separated on a corporate campus or by the miles of airspace between facilities.
If such collaboration grids are widely used in the automotive and aerospace industries, they suggest that the traditional model for Grid 2.0 includes conceptual gaps. In Grid 2.0, grids never move beyond the enterprise, but in collaboration grids they include not only widespread parts of the enterprise, but also a large number of suppliers, subcontractors and engineering partners that the final firm (the auto maker) does not own. In addition, collaboration grids indicate that auto firms are far beyond focusing on “virtualization, aggregation and sharing of all compute, storage, network and data resources.” [“Getting Beyond the Compute Grid: The Challenge of 'Grid 2.0'“, GRIDtoday, Sept. 24] Rather, they are using SOAs and grids to transform the way they bring new products to market, shifting from a technology focus to a market or business focus. And this is where SOAs and grids provide real power — enabling some of the fleetest auto firms to leave their competitors in the dust when it comes to profits and creating cars that are extremely responsive to customer needs.
While much of the recent news in SOA and Grid has been about “services as software,” the virtual car shows that grids are contributing to a major shift in the way goods are produced. If collaboration grids are used across a range of industries, they could exemplify the real power of Grid to alter how firms engage in engineering design. Even service firms might benefit from the creation of a new tool that could shorten the time to design and deliver new services.
Virtual products, not the virtualization of resources, could become one way in which grids change the way the world operates.
About Robert Cohen
Robert B. Cohen is the president of Cohen Communications Group in New York, a fellow with the Economic Strategy Institute and an area director for Industrial Applications of the Open Grid Forum. He is the author of several studies on Grid adoption, including “Grid Computing: Projected Impact on North Carolina’s Economy & Broadband Use through 2010,” and “Cluster and Grid Computing in Japan: Today and in 2010.” Dr. Cohen gave the keynote address at the GRIDtoday '04 conference. He can be reached at [email protected].
Cohen also is one of the coordinators of the OGF-ITU Workshop on Next-Generation Networks and Grids, which is taking place today and tomorrow in Geneva, Switzerland. You can find out more about this workshop at www.itu.int/ITU-T/worksem/grid/programme.html.