Realizing the next generation of supercomputers is here. As the demand for faster, more effective, more energy-saving computer clusters is gaining momentum, achieving impressively high-performance for everything from Google searches to news casts to weather forecasts requires an ever-increasing necessity for more sophisticated computing resources.
The number of processors per system is in the millions and is projected to expand, substantially, in the future; but with that being said, an unchanged, 20-year old element has become stagnant; and that element is the programming model for supercomputers—the MPI. The MPI, or Message Passing Interface, makes sure the microprocessors can communicate. Parallel software has been mainly based on the MPI standard, which was established in 1994 and has dominated applications development. According to researchers in the field of programming, the MPI has reached its limits; and because of that impasse, development has forged ahead with a new programming model.
A New Programming Model:
One researcher at the Fraunhofer Institute for Industrial Mathematics in Germany said this: “Existing models aren’t working and the problems are a lack of scalability, the restriction to bulk-synchronous two-sided communication, and the lack of fault tolerance.” As a result, a new programming model has been developed which has resulted in what is called the GPI, or Global Address Space Programming Interface.
The GPI is based on an absolutely new approach where each processor can directly access all data regardless of which memory it is on without affecting other parallel processes. As with the programming model of MPI, GPI is not a parallel programming language but a parallel programming interface that can be universally utilized. The all-out demand for this type of scalable, flexible and fault-tolerant interface is growing by leaps and bounds. You can fulfill any demand you might have at DMS System, Inc.
Implementing GPI Has Worked:
Implementing GPI has been very successful with industrial applications and this high-performance computing has made itself known as a universal tool in science and business and in fields such as automotive and aircraft manufacturing. As one German researcher put it: “Take the example of aerodynamics: one of the simulation corner-stones in the European aerospace sector, the software TAU, was ported to the GPI platform in a project with the German Aerospace Center. GPI allowed us to significantly increase parallel efficiency.”
Currently, GPI is a tool for specialists, but scientists are convinced it has the potential to transform algorithmic development for high-performance software. And get this—it is considered a crucial component in enabling the next generation of supercomputers to be 1,000 times faster than the mainframes of today! The GPI provides a simple, reliable runtime system to handle large datasets, dynamic and irregular applications that are I/O and compute intensive. All in all, GPI offers an efficient, robust and scalable programming model which continues to be used in many industrial projects and has completely replaced the use of MPI within those projects.
Development Continues:
GPI has been continually evolving and improving since 2005; and GPI-2 is the next generation of GPI with more features including high-performance, flexible API, failure tolerance, memory segments to support heterogeneous systems (e.g. Intel Xeon Phi) and threaded-model and thread-safe interface.
The GPI-2 version has the following requirements—hardware: Infiniband or RoCE devices plus software: OFED software stack installation (in particular libibverbs) and ssh server running on compute nodes (requiring no password).
GPI-2, also, provides atomic operations where variables can be manipulated atomically. There are two main atomic operations: 1) fetch-and-add and 2) compare-and-swap. The values can be used as global-shared variables and to synchronize processes and events.
As the assistant office manager for a search engine marketing business, D. Holdeman functions as a guest poster as a way to support cooperatives inside the USA. He resides in the West Coast, and is soaking up these days with his godly spouse and 3 tykes. Daniel-son calls sightseers to preview his G Plus society when they get a chance.
