# Decentralized Optimal Power Pricing: The Development of a Parallel Program

### Steven Lumetta, Liam Murphy, Xiaoye Li, David E. Culler and Ismail Khalil

###
EECS Department

University of California, Berkeley

Technical Report No. UCB/CSD-93-774

1993

### http://www.eecs.berkeley.edu/Pubs/TechRpts/1993/CSD-93-774.pdf

For MPP's to solve new and interesting problems, they must support the development of sophisticated algorithms on very large data sets. Successful development depends strongly on the speed of the execute-fix cycle. Sequential machines cannot provide sufficiently fast execution of large problems, but many programming systems available on MPP's today neglect the significance of time spent fixing an algorithm during development. Those systems which do address the fix time commonly demand drastic sacrifices in execution speed. Between these two extremes is the middle ground where development must occur. We have implemented a new algorithm to solve an optimization problem for an electrical power system, a problem large enough to require significant computational resources. To help abstract the communication and layout requirements of the problem away from the main algorithm, we have developed a small object system library. The results are an efficient and easily modifiable solution to the problem and a general approach to solving this class of problems.

BibTeX citation:

@techreport{Lumetta:CSD-93-774, Author = {Lumetta, Steven and Murphy, Liam and Li, Xiaoye and Culler, David E. and Khalil, Ismail}, Title = {Decentralized Optimal Power Pricing: The Development of a Parallel Program}, Institution = {EECS Department, University of California, Berkeley}, Year = {1993}, URL = {http://www.eecs.berkeley.edu/Pubs/TechRpts/1993/6295.html}, Number = {UCB/CSD-93-774}, Abstract = {For MPP's to solve new and interesting problems, they must support the development of sophisticated algorithms on very large data sets. Successful development depends strongly on the speed of the execute-fix cycle. Sequential machines cannot provide sufficiently fast execution of large problems, but many programming systems available on MPP's today neglect the significance of time spent fixing an algorithm during development. Those systems which do address the fix time commonly demand drastic sacrifices in execution speed. Between these two extremes is the middle ground where development must occur. We have implemented a new algorithm to solve an optimization problem for an electrical power system, a problem large enough to require significant computational resources. To help abstract the communication and layout requirements of the problem away from the main algorithm, we have developed a small object system library. The results are an efficient and easily modifiable solution to the problem and a general approach to solving this class of problems.} }

EndNote citation:

%0 Report %A Lumetta, Steven %A Murphy, Liam %A Li, Xiaoye %A Culler, David E. %A Khalil, Ismail %T Decentralized Optimal Power Pricing: The Development of a Parallel Program %I EECS Department, University of California, Berkeley %D 1993 %@ UCB/CSD-93-774 %U http://www.eecs.berkeley.edu/Pubs/TechRpts/1993/6295.html %F Lumetta:CSD-93-774