Advanced Computing, Networking
and Storage (ACNS)

As a pathfinder in advanced, distributed, large-scale computing, the ACNS Project tests and evaluates prototype hardware and software to provide guidance to the aeronautics community. These technologies and systems are used as models for implementing similar capabilities within companies, laboratories, or departments. This work is critical to the aerospace community's computing organizations, because it provides essential knowledge for information system investment decisions that no other entity would undertake to acquire on its own. For example, this ACNS research provides critical guidance in provisioning NASA's own production supercomputing facility, the Consolidated Supercomputer Management Office (CoSMO). Additionally, manufacturers and software vendors are given feedback from the testing and evaluation of their "alpha" and "beta" computing and communications technology and products in a strenuous aeronautics production environment. This feedback assists the U.S. computer industry in making their products more suitable for NASA's applications, and more competitive in the world market.

Developing tools and prototypes for a supercomputing metacenter, or "information power grid," is one of the main approaches in the ACNS Project. In this approach, users will submit jobs (demand for supercomputer cycles) to an entire system (grid) of supercomputer, networking, and instrumentation assets. The grid will determine the most appropriate asset, based on capability, cost, and utilization, on which to perform the computation, in much the same way that the source that generates electricity is transparent to the user who plugs a cord into an outlet. The rationale for this approach is that, while progressively higher-performing, less expensive systems continue to be produced, it is clear that a single, monolithic supercomputer system is inappropriate for today's supercomputing needs. If such a system were sized for peak capacity requirements, it would be underutilized most of the time, while if it were sized for average capacity requirements, periods of peak use would suffer inordinate delay. Extraordinarily large systems also tend to be extraordinarily expensive. Furthermore, even when new systems are introduced, organizations must continue to make effective use of their existing infrastructure. The second rationale for this approach is that, while numerical simulations are still growing in demand for resources, the nature of scientific computing has changed to where collaboration, data mining and fusion, and integration with diverse data sources place significant demands on high-end resources. Computing and communications are now pervasive in engineering and scientific research and, therefore, require a distributed, integrated and usable interface to high-end services and resources. The capability for linking a number of systems together in an information grid would solve many of these problems. Existing assets would be effectively utilized to provide capacity to the overall system without creating a bottleneck. Increasing the size of the user base generating demand for capacity would reduce the variability in demand, and hence the difference between average and peak requirements. Large-scale systems built on commodity microprocessors would be used to explore the revolutionary calculations necessary to advance the state of the art in computational simulation, while moderate-sized systems of different architectures would satisfy the basic scientific calculation requirements. Both the large and small systems would leverage off of the commodity microprocessor industry. Finally, the information power grid would offer the efficiencies of consistent, standardized management and the flexibility, failure tolerance, and responsiveness of decentralized operations.

The second objective, partnering with key applications projects, allows the aerospace community early access to state-of-the-art capabilities. Most applications projects use the capability for working on unique problems whose computational requirements in terms of computer speed, storage, and memory exceed those available anywhere else. Furthermore, these users gain the opportunity to assess the performance of their applications codes, which tend to have aerospace-specific computing requirements, on advanced architectures and systems software. The ACNS project strives to balance the computing infrastructure of high-speed computers, hierarchical storage, network communications, and data analysis platforms to maximize the productivity of the system for NASA's largest, most important aerospace applications problems. The ACNS project also provides tools and procedures to assist in solving these problems. They include algorithms, scientific libraries, parallel programming tools and methods, and specialized programs to prepare data for computation to analyze and visualize computational results.

The third objective, pioneering radical new approaches to achieving higher-performance systems, seeks breakthrough technologies to achieve orders-of-magnitude improvements in computing capabilities. While traditional measurements of capability, such as speed, memory, storage, and network bandwidth, continue to be important, the prospect of high-performance computer systems onboard flight and space vehicles elevates the importance of measures such as weight, size, durability, power consumption and thermal management. Miniaturization, device modeling, and architectural research are all components of the project aimed at this objective. Realization of several orders-of-magnitude improvement in computing power and data storage requires exploration of novel devices and nanoelectronics. The goal here is to develop a highly-integrated and intelligent simulation environment that facilitates the rapid development and validation of future generation nanoelectronics devices, as well as associated materials and processes through virtual prototyping at multiple levels of fidelity. Finally, revolutionary software capabilities, such as fault tolerance, transparent computing, legacy code translation, and universal programming models support the overall objective as well.

The Advanced Computing, Storage, and Networking Project consists of three parts designed to address the high performance computing needs of the NASA enterprises:

1. The "information power grid" to integrate a widely distributed and diverse set of computing resources into a secure, useful system
   
   
   
2. Local high-end computing, mass storage systems, networks and infrastructure
   
   
3. Long-term research in future computing devices and systems, such as computational nanotechnology, device modeling and hybrid multithreaded architectures

Part 1, the information power grid, creates the "virtual machine" of services by which a distributed collection of resources can be securely used by problem solving environments, resource management tools, system and application monitoring and distributed applications. Within the information power grid are several individual tasks:

• Problem Solving Environments, High-Level Tools, Programming Services & Reference Applications
  
  
• Grid Common Services
  
  
• High-Performance Resource Management and Security Testbed
  
  
• High Data Rate Testbed
  
  
• IPG Deployment and Operations

Part 2 establishes the local "power" generators, storage, and networks required for computational simulation, data sharing, and remote collaboration.

Part 3 is long-term research toward revolutionary new technologies which will be part of future generations of computing devices required for NASA's research goals. Part 3 has three components: 1) computational nanotechnology, 2) device modeling, and 3) hybrid technology multithreaded architectures.

1. Computational nanotechnology
   
   
2. Device modeling
   
   
3. Hybrid technology multithreaded architecture