New to Parallel Computing?
It might seem silly to have such a small supercomputer without some perspective.
Typical Parallel Computing is Frustrating, Not Fast
Beowulf-class supercomputers (Beowulfs) have dominated the supercomputing and parallel processing scene for the last decade. Beowulfs consist of two or more personal desktop computers connected via network technology and configured with appropriate support software. A typical Beowulf is a collection of hundreds of computers and is maintained in a seperate facility from its users. The machine is almost always shared by many users via a queuing system and maintained by a staff divorced from the purpose of the user. The Message Passing Interface (MPI) is the most common support software for achieving coordination of the program components executing on the machines. The operating system is almost always text-only Linux. These typical conditions often produce a perfect storm of avoidance in most potential users. The most problematic issues caused by the environment are : (1) unpredictable delays imposed by sharing (2) lost or difficult to retrieve output data (3) difficulty of debugging and software development (4) significant added complexity to simply execute a program (5) imposing text-only environment (6) lack of control over system changes (7) drive to keep all processors active degrades parallel execution of one program into execution of many single CPU -- true parallel processing may not even be available. The typical Beowulf experience is inefficient at best and maddening at worst. Baby-B avoids all the typical issues of Beowulf use by providing a Beowulf for your exclusive use! Finally, you are in control your supercomputer!
The Future of Computing is Parallel
Five years ago, one could safely assume that a faster CPU, released in the next few months, would speed up program execution without changing anything about the program. Times have changed, and single CPUs have effectively stopped increasing their execution speed. Now, manufacturers sell CPUs with an ever-increasing number of cores. The old paradigm of assured program speed-up with faster processors no longer holds because the old programs cannot automatically take advantage of the increased number of cores. Now, to acheive speed-up, a program must become parallel. Parallel programs will become ubiquitous for any computationally intense task, even on the desktop of the everyday user. A small supercomputer trumps the performance of one multicore CPU and avoids the inconveniences of "Big Iron" supercomputers.
Speed Up Without Agonizing over Optimization
Our philosophy is simple : Significant execution speed increases do not require optimization. Acheive them with effective parallelization. Baby-B is far cheaper than the work-hours it would take a software engineer to increase the execution speed of a program via painstaking optimization techniques. The time saved with simple parallelization of the program process is well worth the monetary cost. Your time is valuable! The sixteen processors in Baby-B seem small compared to the stats of a "Big Iron" supercomputer, but Baby-B will provides 8-16x the processing power of a typical desktop computer.
Learn Parallel Computing
We have developed an introductory text about parallel computing. It provides everything you need to get started with MPI parallel program in C/C++. Click here to learn more about Supercomputing Simplified: The Bare Necessities of Parallel C Computing with MPI.
