Adding the repository for GPC'2011 conference.

[interreg4.git] / heteropar10 / camera_ready / heteropar10.tex

diff --git a/heteropar10/camera_ready/heteropar10.tex b/heteropar10/camera_ready/heteropar10.tex
index fa91744d36046f96ccace73e2f0f32775aaefccf..f284cc7d9711b700e7036d9f7e49c8fdc604167c 100644 (file)
--- a/heteropar10/camera_ready/heteropar10.tex
+++ b/heteropar10/camera_ready/heteropar10.tex
@@ -24,7 +24,7 @@
 \author{Raphaël Couturier, David Laiymani and Sébastien Miquée}
   \authorrunning{R. Couturier, D. Laiymani and S. Miquée} 
   \institute{\vspace{-0.2cm}
 \author{Raphaël Couturier, David Laiymani and Sébastien Miquée}
   \authorrunning{R. Couturier, D. Laiymani and S. Miquée} 
   \institute{\vspace{-0.2cm}

-     University of Franche-Comté \qquad  LIFC laboratory\\%[1mm]
+     University of Franche-Comté \qquad  LIFC laboratory\\

      IUT Belfort-Montb\'eliard, 2 Rue Engel Gros \\ BP 27 90016 Belfort,
      France\\\{{\tt
       raphael.couturier,david.laiymani,sebastien.miquee}\}{\tt
      IUT Belfort-Montb\'eliard, 2 Rue Engel Gros \\ BP 27 90016 Belfort,
      France\\\{{\tt
       raphael.couturier,david.laiymani,sebastien.miquee}\}{\tt


@@ -116,24 +116,20 @@ algorithms, AIAC-QM (for \textit{AIAC Quick-quality Map}) and F-EC
 (for \textit{Farhat Edges-Cuts}) showed an important performances
 improvement by significantly reducing the application execution
 time. These experiments were performed by using the fully fault
 (for \textit{Farhat Edges-Cuts}) showed an important performances
 improvement by significantly reducing the application execution
 time. These experiments were performed by using the fully fault

-tolerant JaceP2P-V2 environment, described in next section.
-%This Java based platform is an executing and developing
-%environment dedicated to the AIAC model. By implementing a distributed
-%backup/restore mechanism it is also fully fault
-%tolerant\cite{jaceP2P-v2}. 
-In our previous experiments we did not introduce computing nodes
-failures during the computation. As architecture heterogeneity
-continually evolves according to computing nodes volatility, we have
-to take care more precisely about the heterogeneity of the target
-platform. Thus in this paper our main contribution is to propose a new
-mapping algorithm called MAHEVE (\textit{Mapping Algorithm for
-  HEterogeneous and Volatile Environments}).  This algorithm
-explicitly tackles the heterogeneity issue and introduces a level of
-dynamism in order to adapt itself to the fault tolerance
-mechanisms. Our experiments show gains up to $65\%$ on application
-execution time, with faults during executions, which is about 10
-points better than AIAC-QM and about 25 points better than F-EC, and
-MAHEVE also outperforms them in experiments with no fault during executions.
+tolerant JaceP2P-V2 environment, described in next section.  In our
+previous experiments we did not introduce computing nodes failures
+during the computation. As architecture heterogeneity continually
+evolves according to computing nodes volatility, we have to take care
+more precisely about the heterogeneity of the target platform. Thus in
+this paper our main contribution is to propose a new mapping algorithm
+called MAHEVE (\textit{Mapping Algorithm for HEterogeneous and
+  Volatile Environments}).  This algorithm explicitly tackles the
+heterogeneity issue and introduces a level of dynamism in order to
+adapt itself to the fault tolerance mechanisms. Our experiments show
+gains up to $65\%$ on application execution time, with faults during
+executions, which is about 10 points better than AIAC-QM and about 25
+points better than F-EC, and MAHEVE also outperforms them in
+experiments with no fault during executions.

 
 
 The rest of this paper is organized as
 
 
 The rest of this paper is organized as


@@ -200,10 +196,6 @@ on the JaceP2P-V2 platform, interested readers can refer to
 \label{sec:pbmodelapp}
 
 
 \label{sec:pbmodelapp}
 
 

-%With the AIAC model, all tasks compute in parallel at the same time,
-%without precedence nor synchronization. During an iteration, each task
-%computes its job and sends its results to its neighbors, and
-%immediately starts the next iteration. 

 The TIG\cite{tig1} (\textit{Task Interaction Graph}) model is the most
 appropriate to our problem, as it only models relationships between
 tasks. In this model, all the tasks are considered simultaneously
 The TIG\cite{tig1} (\textit{Task Interaction Graph}) model is the most
 appropriate to our problem, as it only models relationships between
 tasks. In this model, all the tasks are considered simultaneously