The MTA overcomes issues that inhibit scaling on
every other system. The MTA is the only system that scales up in
performance without requiring a change in programming model.
The MTA overcomes issues that inhibit scaling on every other system. The MTA is the only system that scales up in performance without requiring a change in programming model.
Vector processors and SMPs approach parallelism from different directions. Vector processors rely on parallel inner loops and a very fast, very expensive memory system. Microprocessors rely on instruction-level parallelism and data caches. As processor speeds increase, vector lengths must become longer and caches must become larger and deeper. In other words, as these processors get faster, they become ever more specialized and less suited for real-world applications.
In an effort to improve scaling, some vendors have abandoned shared memory and introduced distributed-memory computers. These are also euphemistically called scalable parallel, massively parallel, or cluster computers. Regardless of the name, they all suffer the same basic problem: a truly horrible programming model. First, they require that applications be rewritten before they can even be run in parallel. Then, to achieve mediocre levels of performance, they require programs to be carefully tuned to manage communications and data placement. And since these systems are built using off-the-shelf microprocessors, they require further tuning for effective use of their data caches. Finally, these systems all suffer from inadequate communication bandwidth. Parallel applications can never be expected to run as well on these computers as on shared memory systems regardless of the programming effort invested.
Too harsh a critique? Consider . . . the problem with VPPs and SMPs is their poor support for parallelism beyond a single inner loop. MPPs and clusters take the worst feature of the SMPs (the microprocessor), throw out the best feature (shared memory), and complicate matters with non-uniform memory access schemes that make the programmer's life miserable.
We maintain that the MTA's impact on the market will be more profound than that of the Cray 1. Computers like the MTA are required before further real growth can occur in the high-performance computing marketplace.
[Tera's Solution] [A Scalable Computer] [A Programmable Computer] [A Parallel Computer] [A Breakthrough in HPC] [Hardware Characteristics] [Software Characteristics] [Our Conclusion]
[About Tera] [The MTA] [Press Releases] [Investing] [Work at Tera] [Partners] [Library] [Related Sites]