Parallel Processing

To reach higher levels of productivity in the analysis of complex structures with thousands of components, such as in combustion engines or crash simulations, the application of parallel processing was introduced. Two parallel processing methods used in engineering applications are as follows (Fig. 12):

• Symmetric multiprocessing (SMP)

• Massively parallel processing (MPP)

Parallel processing is the use of multiple processors to complete complex tasks and

calculations by dividing up the work. The ultimate goal of parallel processing is to constantly increase the power of the systems. This takes place in a parallel computer or supercomputer

and is similar to having multiple workers work together to complete a single or many tasks in a shorter amount of time. SMP is the shared memory while MPP is the distributed memory used in these processes.

The difference between the SMP and the MPP systems is how they handle the memory relationship with the processor. In an SMP system, the more common of the two, the memory is shared between the two processors. This has the potential to lead to writing problems when both processors attempt to write to the memory at the same time. The MPP system, which is the more powerful of the two, has dedicated memory for each processor and utilizes an interconnection network to deal with the transfer of information from one processor to the next. MPP systems are traditionally more complex than the SMP systems and will usually

be set up in large racks of interconnected processors working on a particular problem. Mostsupercomputer systems today are MPP systems and work on problems such as large simulations or weather forecasting.

MPP systems combine a number of processors into one machine and boast large amounts of processing power. While these machines were expected to revolutionize computerized engineering analysis, users could not simply upload existing applications to these half-billion dollar computers. On the contrary, programming MPP machines proved to be difficult even for experienced programmers, and few applications were ever developed for general use.

SMP, however, has shown extraordinary potential for use in general engineering analysis. SMP techniques essentially link PCs or workstations, each with existing memory and disk storage capacities and CPU capabilities, allowing the attributes of each individual machine to be applied to the computations. In addition, there is no lack of application software for SMP users. Applications for thermal, dynamic, stress (both linear and nonlinear), and fluid analyses are available for use with SMP computers. To meet its needs for a more powerful supercomputer, the U.S. Department of Energy ran a project called ASCI (Accelerated Strategic Computing Initiative) which would require computational speeds of 100 TFLOPS (one trillion floating points per second, also known as teraflop). This project started in 1995 and had its first machine built for the project at the Sandia National Laboratory. This machine, called the ASCI Option Red Supercomputer, ran 9152 Pentium Pro processors and was the first computer to run more than 1 TFLOPS on a real application.

Another variation of parallel processing is through the use of PCs connected to the Internet. The  ETI@Home project, for example, has its volunteer users download software that will run when the primary users are not using it. SETI (Search for Extraterrestrial Intelligence) analyzes radio telescope data in an effort to detect signals of intelligent beings from outer space.

READ MORE.......

Emory W. Zimmers, Jr. and Technical Staff

Enterprise Systems Center

Lehigh University

Bethlehem, Pennsylvania

for STEP BY STEP GUIDE unigraphics simple tutorial please visit.........

www.unigraphicsimpletutorial.blogspot.com

---or---

www.unigraphic-simple-tutorial.com

---or---

www.solidwork-simple-tutorial.com

Berbagi :