CarpetIOStreamedHDF5: introduce a parameter max_num_clients to limit the number of clients allowed to connect to a simulation at the same time

darcs-hash:20050701100728-776a0-7964598e966223fa1db8dddfce86502d26992ac4.gz

4 files changed, 29 insertions, 9 deletions

diff --git a/Carpet/CarpetIOStreamedHDF5/doc/documentation.tex b/Carpet/CarpetIOStreamedHDF5/doc/documentation.tex
index e5742ecb1..55721f59d 100644
--- a/Carpet/CarpetIOStreamedHDF5/doc/documentation.tex
+++ b/Carpet/CarpetIOStreamedHDF5/doc/documentation.tex
@@ -64,14 +64,6 @@ any time.
 
 Parameters to control the \ThisThorn\ I/O method are:
 \begin{itemize}
-  \item {\tt IOStreamedHDF5::port}\\
-        The initial port number which should be opened at startup for
-        client connections. If the given port is already in use by some other
-        program, {\tt IOStreamedHDF5} will search for the next available port
-        by advancing {\tt IOStreamedHDF5::port}.\\
-        The actual port number used by \ThisThorn\ will be output in a INFO
-        message at startup, along with the hostname of the root processor.
-
   \item {\tt IOStreamedHDF5::out\_every} (steerable)\\
         How often to do periodic \ThisThorn\ output. If this parameter
         is set in the parameter file, it will override the setting of the shared
@@ -106,6 +98,25 @@ t\_dt})
         whether to output the data from multiple processors in chunked or
         unchunked format
 
+  \item {\tt IOStreamedHDF5::port}\\
+        The initial port number which should be opened at startup for
+        client connections. If the given port is already in use by some other
+        program, {\tt IOStreamedHDF5} will search for the next available port
+        by advancing {\tt IOStreamedHDF5::port}.\\
+        The actual port number used by \ThisThorn\ will be output in a INFO
+        message at startup, along with the hostname of the root processor.
+
+  \item {\tt IOStreamedHDF5::max\_num\_clients}\\
+        The overhead of serving many clients with streamed HDF5 data can slow
+        down your simulation notably. When you are running both the simulation
+        and the visualisation client on the same machine, it could also happen
+        that the client is fast enough to reconnect to the simulation while
+        it is still serving the same timestep -- which leads to duplicate
+        data streaming.\\
+        In order to prevent this, you can set the {\tt max\_num\_clients}
+        parameter to limit the maximum number of clients allowed to connect
+        to the simulation at the same time.
+
 \end{itemize}
 
 
diff --git a/Carpet/CarpetIOStreamedHDF5/par/CarpetIOStreamedHDF5.par b/Carpet/CarpetIOStreamedHDF5/par/CarpetIOStreamedHDF5.par
index 7818242dd..c1f8fd05b 100644
--- a/Carpet/CarpetIOStreamedHDF5/par/CarpetIOStreamedHDF5.par
+++ b/Carpet/CarpetIOStreamedHDF5/par/CarpetIOStreamedHDF5.par
@@ -57,3 +57,7 @@ IO::parfile_write        = "no"
 
 IOBasic::outInfo_vars    = "wavetoy::phi"
 IOStreamedHDF5::out_vars = "wavetoy::phi"
+
+# allow only one client at at time to connect to the simulation
+# (to prevent potentially multiple connects from the same client)
+IOStreamedHDF5::max_num_clients = 1
diff --git a/Carpet/CarpetIOStreamedHDF5/param.ccl b/Carpet/CarpetIOStreamedHDF5/param.ccl
index 6db82cc93..1aacda901 100644
--- a/Carpet/CarpetIOStreamedHDF5/param.ccl
+++ b/Carpet/CarpetIOStreamedHDF5/param.ccl
@@ -37,6 +37,11 @@ INT port "Port number for clients to connect to" STEERABLE = RECOVER
   1000:32000 :: "Ports below 1000 require root access"
 } 10000
 
+INT max_num_clients "Maximum number of clients allowed to connect at the same time" STEERABLE = RECOVER
+{
+  1:20 :: "a small positive number"
+} 3
+
 
 #############################################################################
 ### import IOUtil parameters
diff --git a/Carpet/CarpetIOStreamedHDF5/src/CarpetIOStreamedHDF5.cc b/Carpet/CarpetIOStreamedHDF5/src/CarpetIOStreamedHDF5.cc
index 65e8ccdbf..92a47ff5a 100644
--- a/Carpet/CarpetIOStreamedHDF5/src/CarpetIOStreamedHDF5.cc
+++ b/Carpet/CarpetIOStreamedHDF5/src/CarpetIOStreamedHDF5.cc
@@ -266,7 +266,7 @@ static int OutputGH (const cGH* const cctkGH)
         fapl.increment = 0;
         fapl.socket = myGH->socket;
         fapl.do_socket_io = 1;
-        fapl.backlog = 5;
+        fapl.backlog = max_num_clients;
         fapl.broadcast_fn = NULL;
         fapl.broadcast_arg = NULL;