A few updates.

diff --git a/supercomp11/biblio.bib b/supercomp11/biblio.bib
index 70a6f87c755af9c2f067775da1031be37415b289..5e721a15fed4d864dc5479713f4672ff083e495e 100644 (file)
--- a/supercomp11/biblio.bib
+++ b/supercomp11/biblio.bib
@@ -1,3 +1,20 @@
+@Article{10.1109/TPDS.2005.2,
+  author =       {Jacques M. Bahi and Sylvain Contassot-Vivier and
+                  Raphael Couturier and Flavien Vernier},
+  title =        {A Decentralized Convergence Detection Algorithm for
+                  Asynchronous Parallel Iterative Algorithms},
+  journal =      {IEEE Transactions on Parallel and Distributed
+                  Systems},
+  volume =       {16},
+  number =       {1},
+  issn =         {1045-9219},
+  year =         {2005},
+  pages =        {4-13},
+  doi =          {10.1109/TPDS.2005.2},
+  publisher =    {IEEE Computer Society},
+  address =      {Los Alamitos, CA, USA},
+}
+
 @InProceedings{bahi+giersch+makhoul.2008.scalable,
   author =       {Jacques M. Bahi and Arnaud Giersch and Abdallah
                   Makhoul},

diff --git a/supercomp11/supercomp11.tex b/supercomp11/supercomp11.tex
index 01ca4f7f4f1a7df8840ec538cef759991324d609..056e185e70929d8d7b38059a671468b5f0b7712c 100644 (file)
--- a/supercomp11/supercomp11.tex
+++ b/supercomp11/supercomp11.tex
@@ -489,8 +489,18 @@ To summarize the different load balancing strategies, we have:
 %
 This gives us as many as $4\times 2\times 2 = 16$ different strategies.
 
+\paragraph{End of the simulation}
 
-\paragraph{Configurations}
+The simulations were run until the load was nearly balanced among the
+participating nodes.  More precisely the simulation stops when each node holds
+an amount of load at less than 1\% of the load average, during an arbitrary
+number of computing iterations (2000 in our case).
+
+Note that this convergence detection was implemented in a centralized manner.
+This is easy to do within the simulator, but it's obviously not realistic.  In
+a real application we would have chosen a decentralized convergence detection algorithm, like the one described in \cite{10.1109/TPDS.2005.2}.
+
+\paragraph{Platforms}
 
 In order to show the behavior of the different strategies in different
 settings, we simulated the executions on two sorts of platforms.  These two
@@ -511,6 +521,8 @@ bandwidth was fixed to 2.25~GB/s, with a latency of 500~$\mu$s.
 Then we derived each sort of platform with four different number of computing
 nodes: 16, 64, 256, and 1024 nodes.
 
+\paragraph{Configurations}
+
 The distributed processes of the application were then logically organized along
 three possible topologies: a line, a torus or an hypercube.  We ran tests where
 the total load was initially on an only node (at one end for the line topology),
@@ -552,12 +564,16 @@ time.
 
 \paragraph{Metrics}
 
+In order to evaluate and compare the different load balancing strategies, we
+choose to measure the following metrics:
+
 \begin{description}
-\item[\textbf{average idle time}]
-\item[\textbf{average convergence date}]
-\item[\textbf{maximum convergence date}]
-\item[\textbf{data transfer amount}] relative to the total data amount
+\item[\textbf{average idle time:}]
+\item[\textbf{average convergence date:}]
+\item[\textbf{maximum convergence date:}]
+\item[\textbf{data transfer amount:}] relative to the total data amount
 \end{description}
+\FIXME{dire à chaque fois ce que ça représente, et motiver le choix}
 
 \subsection{Validation of our approaches}
 \label{Results}