Resumen
Uniting commodity processors by thousands, emerging parallel architectures promise solutions to complex calculations, at least until laws of physics impose a fundamental barrier to human innovation. The demand for computing power is not just for processing speed but also for memory size and speed and data input and output rate as well. As the volume of sensor and simulation output data increases, large archival storage systems with rapid data retrieval play increasingly important role. Scalable parallel processors (SPP) are systems that can be configured with hundreds or even thousands of state-of-the-art processors. The challenge for SPP is to increase not so much the number of processors but the logical and physical interconnections, allowing a large number to be used effectively as one computing system. The challenge for designers of SPP systems is to keep these complex processors fed with data so they can perform close to their potential. INSETS: Promise and Limits of Amdahl's Law and Moore's Law; Reflecting the Trends; SIMD vs. MIMD vs. VLIW.