TU Delft Algorithmics
Jan Renze Steenhuisen
Delft University of Technology Projects ALG Group
EWI ALG Renze SteenhuisenProjects
eng
 
 
 
 
 
 
 
 
 
 
Service Coordination in Grids
Project's website
This workpackage is part of a large project called ICIS. In this project, research is done in the field of actor-agent communities where humans (actors) and software (agents) work together to achieve a common goal. Although these actor-agent communities can come to existence in many real-life scenario's, the crisis domain is taken as the demonstration domain for the developed theory and software tools. The goal for the workpackage service coordination in grids is to come up with coordination mechanisms such that the different services that are available in a grid are guaranteed to give a certain performance characteristic. Previous work on coordination has considered the prevention of deadlock. We continue on this work, but expansions are needed to be a serious candidate to use within a crisis scenario (e.g., timely-performance guarantees, reasoning with a wide variety of temporal constraints).

Project\'s website'); begin_text(); ?> This project was started to restore the complete set of 3, 4, 5, and 6-men EndGame TableBases. The total amount of this collection is approximately 1 TeraByte. Besides personal contacts and direct sharing through ftp, a peer-to-peer system is used to share the available EGTBs to everyone who is interested. An important goal of the project is to have a robuust backup system, such that this large amount of data is not going to be lost again or only partially available to the public. Data consistency is of the utmost importance, and needs to be verified by either its md5sum (e.g., with md5sum.exe) or by using the datacomp.exe program.

EGTBs

TBS files

MD5sums

21, 21p tbs-3-men.zip 21, 21p
22, 22p, 31, 31p tbs-4-men.zip 22, 22p, 31, 31p
32, 32p, 41, 41p tbs-5-men.zip 32, 32p, 41, 41p
33, 33p, 42, 42p, 51, 51p tbs-6-men.zip 33, 33p, 42, 42p
In this project, I use chess programs to search certain sets of positions very deep (i.e., search depths of at least 18 ply). With this data, many interesting questions can be answered, and statistically significant when the amount of data is large enough. On such interesting question is: ``How does the best-change rate (i.e., change that the best move change at current iteration differs from the previous one) differ for increasing search depths?'' Similar questions can be formulated and studied as well (see New Results in Deep-Search Behaviour for a good overview). In summary, the following sets of positions were analysed: Bratko-Kopec (20 ply, 10 hours), Best-Ways to Checkmate (20 ply, 10 hours), Crafty-343 (20 ply), Encyclopedia of Chess Middlegames (18 ply), Encyclopedia of Chess Openings (18 ply), Maastricht Test Suite (20 ply, 20 hours), MES (20 ply, 10 hours), Nolot (20 ply, 20 hours), and Win at Chess (20 ply, 20 hours).
CABS
Collective Agent-Based Systems
ICIS