Correction and re-formatting.

[interreg4.git] / pdsec2010 / pdsec2010.tex

diff --git a/pdsec2010/pdsec2010.tex b/pdsec2010/pdsec2010.tex
index eb27bdfbd7ddd1ecae522cb61fd9417b98c046f5..e363475cf1087b10276b489ee4114c4c9d6685db 100644 (file)
--- a/pdsec2010/pdsec2010.tex
+++ b/pdsec2010/pdsec2010.tex
@@ -823,7 +823,10 @@ this model, the faster and more frequently a task receives its
 dependencies, the faster it converges. Moreover, as the JaceP2P-V2
 environment is fault tolerant and tasks save checkpoints on their
 neighbors, it is more efficient to save on near nodes than on far
-ones.
+ones. In the synchronous model, both heterogeneity and locality must
+be taken into account in a balanced way. In the asynchronous model,
+since no synchronization occurs, the heterogeneity issue is less
+important.
 
 
 %As our problem is on task mapping, only considering the dependency
@@ -1421,7 +1424,7 @@ in comparison to the version without mapping.
   \caption{Gains in time of the execution of the class E of the CG 
     application on Arc1.1 using 64 nodes}
   \label{tab:exph1E}
-  \vspace*{-0.3cm}
+  \vspace*{-0.4cm}
 \end{table}
 
 %\vspace{-0.5cm}
@@ -1444,7 +1447,7 @@ in comparison to the version without mapping.
   \caption{Gains in time of the execution of the class F of the CG
     application on Arc1.2 using 128 nodes}
   \label{tab:exph1F}
-  \vspace*{-0.3cm}
+  \vspace*{-0.4cm}
 \end{table}
 
 At first, we can see that the Simple Mapping algorithm, though it is
@@ -1509,7 +1512,7 @@ mapping.
   \caption{Gains in time of the execution of the class E of the CG
     application on Arc2.1 using 64 nodes}
   \label{tab:exph2E}
-  \vspace*{-0.3cm}
+  \vspace*{-0.4cm}
 \end{table}
 
 \renewcommand{\arraystretch}{1.5}
@@ -1530,7 +1533,7 @@ mapping.
   \caption{Gains in time of the execution of the class F of the CG
     application on Arc2.2 using 128 nodes}
   \label{tab:exph2F}
-  \vspace*{-0.3cm}
+  \vspace*{-0.4cm}
 \end{table}
 
 To begin with, these experiments confirm that a mapping algorithm is
@@ -1743,7 +1746,7 @@ a whole cluster fails, as we save on neighbors (which are
 in general on the same cluster for communication efficiency reasons),
 an important part of the application is lost and we cannot restart
 this part; so the whole application fails. A trade-off should be done
-by having some saving nodes in external clusters.
+by having some saving nodes in external clusters.\\
 
 % \subsubsection*{Acknowledgements}