Level-1 Miletone 350

Quinn, T

Lawrence Livermore National Laboratory

关键词: Management;    Storage;    99 General And Miscellaneous//Mathematics, Computing, And Information Science;    Computers;    Functionals;   

DOI  :  10.2172/892075
RP-ID  :  UCRL-TR-222521
RP-ID  :  W-7405-ENG-48
RP-ID  :  892075

美国|英语

来源: UNT Digital Library

【 摘 要 】

This milestone is the direct result of work that started seven years ago with the planning for a 100 Teraflop platform and will be satisfied when 100 Teraflops is placed in operation and readied for Stockpile Stewardship Program simulations. The end product of this milestone, code named Purple, will be a production high-performance computing system designed to be used to solve the most demanding stockpile stewardship problems, that is, the large-scale application problems at the edge of our understanding of weapon physics. This fully functional 100 TeraOP/s system must be able to serve a diverse scientific and engineering workload. It must also have a robust code development and production environment both of which facilitate the workload requirements. This multi-year effort includes major activities in contract management, facilities, infrastructure, system software, and user environment and support. Led by LLNL, the Tri-labs defined the statement of work for a 100 Teraflop system that resulted in a contract with IBM known as the Purple contract. LLNL worked with IBM throughout the contract period to resolve issues and collaborated with the Program to resolve contractual issues to ensure delivery of a platform that best serves the Program for a reasonable cost. The Purple system represents a substantial increase in the classified computer resources at LLNL for NNSA. The center computer environment must be designed to accept the Purple system and to scale with the increase of compute resources to achieve required end-to-end services. Networking, archival storage, visualization servers, global file systems, and system software will all be enhanced to support Purple's size and architecture. IBM and LLNL are sharing responsibility for Purple's system software. LLNL is responsible for the scheduler, resource manager, and some code development tools. Through the Purple contract IBM is responsible for the remainder of the system software including the operating system, parallel file system, and runtime environment. LLNL, LANL, and SNL share responsibility for the Purple user environment. Since LLNL is the host for Purple, LLNL has the greatest responsibility. LLNL will provide customer support for Purple to the Tri-labs and as such has the lead for user documentation, negotiating the Purple Usage Model, mapping of the ASC Computational Environment requirements to the Purple environment and demonstrating those requirements have been met. In addition LLNL will demonstrate important capabilities of the computing environment including full functionality of visualization tools, file transport between Purple and remote site file systems and the build environment for principle ASC codes. LANL and SNL are responsible for delivering unique capabilities in support of their users, port important applications and libraries, and demonstrating remote capabilities. The key capabilities that LANL and SNL will test are user authorization and authentication, data transfer, file system, data management, and visualization. SNL and LANL should port and run in production mode a few key applications on a substantial number of Purple nodes.