High Performance Computing involves the delivery of computing as a service rather than a product, whereby shared resources, software and information are provided to computers and other devices as a utility (like the electrical grid) over a network (typically the internet). It fills the need to increase capacity or add capabilities on the fly without investing in new infrastructure, training new personnel, or licensing new software.
We work with providers to deliver applications accessed from a web browser, while the business software and data are stored on servers at a remote location. LT can also assist with the building or improvement of data centers, and more effective implementation of SLAs (service-level agreements). LT also works with distributed, grid, peer-to-peer, utility, and service-oriented computing, sometimes referred to as “sofware as a service.”
Massively Parallel Processing
The future of computing has been stalled by the limit of how fast we can operate a processor. For the past few years the “speed wall” has held, with CPUs limited to less than 4 GHz. In order for computers to continue exponential increases in power, we must move to massively parallel systems; many processors operating together to solve a single problem.
Shifting to massively parallel systems means not only a change in hardware design, but a complete overhaul of both software architecture and the way software is designed and developed. Today’s computing hardware, software, and software building tools were designed decades ago for systems with one or a small number of cores..
Today’s operating systems, designed decades ago, do not support or are not efficient for massively parallel computing.
Unix (Apple Mac OSX, Linux) and Windows architectures were designed for single core systems. Later adaptations for multi-core processors simply patched the architecture to support a small number of cores, but they quickly become inefficient above 6 or 8 cores. In addition, the programming environments for these operating systems make the planning and use of multiple cores difficult, leading to software which can not take advantage of massively parallel power.
Many Integrated Core Architecture
Intel introduced its “Many Integrated Core” architecture, or MIC, last year. Their first MIC processor contained 64 CPUs on a single chip. Surprisingly, these chips are the same size and consume the same amount of power as Intel’s current top-of-the-line desktop CPUs. Future MIC processors may contain 1,000 CPUs or more on a single chip, and will be no larger nor require more power than today’s high-end desktop computers.
Intel’s OS solution for their MIC platform with 64 cores is to run 64 individual instances of Linux; treating each integrated core as a separate mini-computer. Not only is this an incredibly inefficient use of processing power and memory, it does not allow information to be easily and quickly shared between cores. There is simply no OS solution on the market for massively parallel systems.
Parallel OS is Leveraging Technologies’ proprietary massively parallel operating system. The architecture, the kernel model, and even the development environment are designed from the ground-up for massively parallel systems. Scalable to over 2 million cores, Parallel OS is the only operating system capable of leveraging the power inherent in massively parallel processors like Intel’s MIC platform. It is the only current commercial operating system that is future-proof, designed for the only processor paradigm that allows for future increases in processing power.
Parallel OS Today: Vertical Appliances
Vertical Appliances are single-purpose computers, designed and optimized for specific tasks, such as database searching, computer graphics, or serving web pages.
Leveraging Technologies’ proprietary massively parallel LT Operating System, paired with Intel’s MIC massively parallel hardware architecture, give our vertical appliances unprecedented performance per watt of power, per cubic inch of space, and per dollar.
Data centers can replace existing Unix and Windows server farms and immediately realize an exponential increase in processing capacity, lower power consumption and less heat generation. Additionally, because one Leveraging Technologies’ vertical appliance can replace as many as 8 traditional Windows or Unix servers, a data center can add more capacity with the same floor space, cooling, and power supply, or reduce physical footprint leading to cost savings on physical plants.
For institutions that require high-performance computing, such as medical imaging or data mining, Parallel OS Vertical Appliances mean faster results. With both a OS and Kernel, and development environment designed for massively parallel processing, a single Parallel OS Desktop Vertical Appliance can return results multiple times faster than the fastest Windows or Unix workstations on the market.
Parallel OS Tomorrow: General Purpose Computing
As massively parallel design makes its way into home and office computing, Apple, Microsoft, and the various Unix sources will need to design and build completely new solutions if they wish transition to the future with massively parallel systems. Leveraging Technologies’ Parallel OS will be a proven massively parallel operating system already in the marketplace. With the addition of a consumer-level graphic interface, Parallel OS will offer OEM system suppliers a market-tested, mature platform for their new machines.
Parallel OS in the office means higher quality remote collaboration, fewer and faster servers, lower cost of operations, and faster results.
Parallel OS in the home means centralized personal computing, more realistic video games, and lower cost support for high quality video platforms like UltraHD.
Parallel OS Overview
Parallel OS is a completely new model for a commercial operating system and unlike anything else on the market. It is designed from the ground up to leverage the raw power of Intel’s massively integrated core CPUs, something impossible for any OS in the marketplace today.
Designed for CPU architectures with a high core count
- Today: Intel MIC “Knights Bridge” CPU with 72 cores
- Tomorrow: Intel future CPU with 1000+ cores
High-performance Kernel and OS:
- 64 bit
- Based on Exokernel architecture model
Core support for parallel programming models:
- Designed at the lowest levels to fully support concurrent software design and execution.
- Support for traditional languages with parallel extensions: C/C++ with Intel Clik Plus, Join Java
- Support for newer highly concurrent languages: Erlang, Haskell, Clojure
- Today: Vertical Appliances
- Tomorrow: Generic Appliances
- Future: Generic OS