New version of MAHEVE plus corrections. - New version of MAHEVE: performance enhancement. - Adding some functionalities in the library for the new version of MAHEVE and to enhance library performance.
Adding new functionalities. Adding new functionalities to improve the Maheve algorithm. Correction of some functions in order to take into consideration multiples executions (Utils, Grid).
Improvement of the Maheve algorithm. Improvement by adding the possibility to change the parameters "nbSave" and "minNode" to adjust the mapping process. Also a correction in the mapping process, especially in the nodes search, to avoid an error when using clusters with few nodes; with an amount of nodes near the "minNode" limit.
New stable version with the add of the Maheve algorithm. - Adding the new efficient MAHEVE mapping algorithm. - Correction of some bugs. - Adding minor functionalities. - Cleaning code.
Implementation of new functions. - Adding some functions in Cluster in order to compute new indicators for a novel mapping algorithm. - Adding new functions in Utils to read and write Mapping objects in a file. - Returning to the real relative standard deviation for heterogeneity degree computation in Grid.
Separate the javadoc directory of the project sources.
Implementation of fault tolerance functions in all algorithms. - In DefaultMapping, it returns in both cases the next node in the list. - In QM, it returns the best computing power node for the replacement, and a node from the cluster which has the most amount of free nodes for another node. - In LSM (Edge-cuts), it returns the nearest node for the replacement, and a node from the cluster which has the most amount of free nodes for another node. - Simple does the same thing as DefaultMapping.
Adding some new functions. - Adding of functions to facilitate the (re)initialization of nodes' status. -Correction of some bugs.
Correction of bugs and new functionalities. - Simplification of some functions and suppression of unused or not useful variables. - Correction of some functions after having done different tests of non regression.
Correction of a bug. - In Mapping class, an error was encountered in the add of an Association in the mapping.
Implementation of fault tolerance in Default Mapping. - Implementing the "replaceNode" and "getOtherNode" in the Default Mapping algorithm. - Adding a new function in the Mapping class and two others in the Association class in order to facilitate the replacing of a dead node.
Purge of Javadoc.
Many modifications in order to make a more efficient library for GNode creation and graph management. 1) The xtream jar was uncompressed in order to use classes directly (Java does not include jar within another jar) 2) The sigar library was deleted. Indeed, some problems with its use were difficult to solve and another more important problem was found. With this library we obtain system information, like processor frequency or amount of memory, but this not the real information we need. Indeed, as we execute our program in a JVM, we should only take into consideration what the JVM provide to us. So all information retrieved are information about the JVM, which are: amount of available computing cores and memory and the MegaFlops of a computing core. 3) In the grid class was added the computation of the heterogeneity degree of the plate-form. This would be useful in designing a mapping algorithm which takes care about this parameter.