Скачать презентацию A Migration Framework for Legacy Scientific Applications w Скачать презентацию A Migration Framework for Legacy Scientific Applications w

14a054d10c109a70b32711e9cbe3f0fa.ppt

  • Количество слайдов: 7

A Migration Framework for Legacy Scientific Applications w Current tendency: n n n monolithic A Migration Framework for Legacy Scientific Applications w Current tendency: n n n monolithic architectures large, complex, tightly coupled implementations ad-hoc evolution over time w Downside? n n limited number of developers who can modify/extend embedded scientific knowledge accrued is trapped w Upside? n n rich and valuable science job security? 1/30/2003 Los Alamos National Laboratory 1

Desirable Features w An architecture that expands the developer base w Retain the rich Desirable Features w An architecture that expands the developer base w Retain the rich scientific content w Modular, component-based architecture to enable a large, but manageable feature set 1/30/2003 Los Alamos National Laboratory 2

Desirable Outcomes w A system architecture with industry standard form approachable by a large Desirable Outcomes w A system architecture with industry standard form approachable by a large audience w Re-architect with an OO infrastructure to retain embedded science w Establish well-defined interfaces to support n n generalized input mechanisms multiple plug-able physics models generalized output mechanisms integration of independently developed packages 1/30/2003 Los Alamos National Laboratory 3

Tools and Methods w Use accepted software engineering practices n Source code version control Tools and Methods w Use accepted software engineering practices n Source code version control n Configuration management n Automated regression testing n Multiple platform portability w Use standards as much as possible n CORBA (components) n Patterns (modules) n Languages (objects) See Los Alamos National Laboratory, LA-UR 01 -6629 1/30/2003 Los Alamos National Laboratory 4

Results to Date w Trust established with automated regression testing w Optimized CORBA IDL Results to Date w Trust established with automated regression testing w Optimized CORBA IDL implementation success on several UNIX platforms based on dynamic libraries n n Linux Solaris HP/UX IRIX w Modules in multiple languages (on Linux) n n n Fortran 90 C C++ w Priority changes resulting in loss of interest and funding 1/30/2003 Los Alamos National Laboratory 5

Future Efforts and Approaches w Find interest and funding w Try CHASM (language glue) Future Efforts and Approaches w Find interest and funding w Try CHASM (language glue) See http: //acts. nersc. gov/presentations/Workshop 2002/ slides/Rasmussen. pdf w Write IDL compiler for support 1/30/2003 Los Alamos National Laboratory 6

1/30/2003 Los Alamos National Laboratory 7 1/30/2003 Los Alamos National Laboratory 7