diff --git a/apps/benchmarks/CMakeLists.txt b/apps/benchmarks/CMakeLists.txt
index 2e835d671e25837aad63e89e6534f9e44ade2384..d5612a35b72b52f7c2bf1d5e0cb98da6dfbd1934 100644
--- a/apps/benchmarks/CMakeLists.txt
+++ b/apps/benchmarks/CMakeLists.txt
@@ -8,6 +8,7 @@ add_subdirectory( NonUniformGrid )
 add_subdirectory( MotionSingleHeavySphere )
 add_subdirectory( PeriodicGranularGas )
 add_subdirectory( PoiseuilleChannel )
+add_subdirectory( ProbeVsExtraMessage )
 add_subdirectory( SchaeferTurek )
 add_subdirectory( UniformGrid )
 add_subdirectory( UniformGridGPU )
diff --git a/apps/benchmarks/ProbeVsExtraMessage/CMakeLists.txt b/apps/benchmarks/ProbeVsExtraMessage/CMakeLists.txt
new file mode 100644
index 0000000000000000000000000000000000000000..7330aa0ddb83f973d47071a4dba1a4c03baed242
--- /dev/null
+++ b/apps/benchmarks/ProbeVsExtraMessage/CMakeLists.txt
@@ -0,0 +1,2 @@
+waLBerla_add_executable ( NAME ProbeVsExtraMessage 
+                          DEPENDS core postprocessing stencil )
diff --git a/apps/benchmarks/ProbeVsExtraMessage/ProbeVsExtraMessage.cpp b/apps/benchmarks/ProbeVsExtraMessage/ProbeVsExtraMessage.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..6fec0642ef3a5f476955403eaa3bd77b22fe96ea
--- /dev/null
+++ b/apps/benchmarks/ProbeVsExtraMessage/ProbeVsExtraMessage.cpp
@@ -0,0 +1,211 @@
+//======================================================================================================================
+//
+//  This file is part of waLBerla. waLBerla is free software: you can
+//  redistribute it and/or modify it under the terms of the GNU General Public
+//  License as published by the Free Software Foundation, either version 3 of
+//  the License, or (at your option) any later version.
+//
+//  waLBerla is distributed in the hope that it will be useful, but WITHOUT
+//  ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+//  FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+//  for more details.
+//
+//  You should have received a copy of the GNU General Public License along
+//  with waLBerla (see COPYING.txt). If not, see <http://www.gnu.org/licenses/>.
+//
+//! \file ProbeVsExtraMessage.h
+//! \author Martin Bauer <martin.bauer@fau.de>
+//! \author Sebastian Eibl <sebastian.eibl@fau.de>
+//! \brief Micro Benchmark, measuring time for different variable sized communications
+//
+//======================================================================================================================
+
+#include "core/DataTypes.h"
+#include "core/Environment.h"
+#include "core/math/Vector3.h"
+#include "core/mpi/BufferSystem.h"
+#include "core/mpi/MPIManager.h"
+#include "core/timing/TimingPool.h"
+#include "postprocessing/sqlite/SQLite.h"
+#include "stencil/D3Q27.h"
+#include "stencil/D3Q19.h"
+#include "stencil/D3Q7.h"
+
+#include <array>
+#include <iostream>
+#include <sstream>
+
+namespace walberla {
+
+class MPIInfo
+{
+public:
+   MPIInfo( const Vector3<uint_t>& procs, const Vector3<bool>& periodicity );
+
+   int getNeighborRank(const stencil::Direction& dir);
+private:
+   Vector3<uint_t> procs_;
+   Vector3<bool>   periodicity_;
+   Vector3<int>    pos_;
+};
+
+MPIInfo::MPIInfo( const Vector3<uint_t>& procs, const Vector3<bool>& periodicity )
+   : procs_(procs)
+   , periodicity_(periodicity)
+{
+   mpi::MPIManager::instance()->createCartesianComm(procs[0], procs[1], procs[2], periodicity[0], periodicity[1], periodicity[2]);
+   mpi::MPIManager::instance()->cartesianCoord(pos_.data());
+}
+
+int MPIInfo::getNeighborRank( const stencil::Direction& dir )
+{
+   auto neighborCoord = pos_;
+   neighborCoord[0] += stencil::cx[dir];
+   neighborCoord[1] += stencil::cy[dir];
+   neighborCoord[2] += stencil::cz[dir];
+   if (neighborCoord[0] < 0) return -1;
+   if (neighborCoord[1] < 0) return -1;
+   if (neighborCoord[2] < 0) return -1;
+   if (neighborCoord[0] >= int_c(procs_[0])) return -1;
+   if (neighborCoord[1] >= int_c(procs_[1])) return -1;
+   if (neighborCoord[2] >= int_c(procs_[2])) return -1;
+   return mpi::MPIManager::instance()->cartesianRank(uint_c(neighborCoord[0]), uint_c(neighborCoord[1]), uint_c(neighborCoord[2]));
+}
+
+template <typename Stencil>
+void communicate( MPIInfo& mpiInfo,
+                  const uint_t iterations,
+                  const uint_t messageSize,
+                  const bool iProbe,
+                  WcTimingPool& tp)
+{
+   std::vector<char> sendBuf(messageSize);
+   std::vector<char> recvBuf(messageSize);
+
+   WcTimer& timer = tp[iProbe ? "IProbe" : "twoMessage"];
+
+   mpi::BufferSystem bs( mpi::MPIManager::instance()->comm() );
+   bs.useIProbe(iProbe);
+
+   for( uint_t i =0; i < iterations; ++i )
+   {
+      timer.start();
+
+      for (auto dirIt = Stencil::beginNoCenter(); dirIt != Stencil::end(); ++dirIt)
+      {
+         auto recvRank = mpiInfo.getNeighborRank( *dirIt );
+         if (recvRank == -1) continue;
+         bs.sendBuffer(recvRank) << sendBuf;
+         WALBERLA_ASSERT_EQUAL(bs.sendBuffer(recvRank).size(), messageSize + sizeof(size_t));
+      }
+
+      bs.setReceiverInfoFromSendBufferState(false, true);
+      bs.sendAll();
+
+      for( auto it = bs.begin(); it != bs.end(); ++it )
+      {
+         WALBERLA_ASSERT_EQUAL(it.buffer().size(), messageSize + sizeof(size_t));
+         it.buffer() >> recvBuf;
+         WALBERLA_ASSERT_EQUAL(recvBuf.size(), messageSize);
+         WALBERLA_ASSERT(it.buffer().isEmpty());
+      }
+
+      timer.end();
+   }
+}
+
+int main( int argc, char ** argv )
+{
+   mpi::Environment mpiEnv( argc, argv );
+
+   if ( argc != 10 )
+   {
+      WALBERLA_ROOT_SECTION()
+      {
+         std::cout << "Usage ./probeVsExtraMessage x y z px py pz iterations messageSize stencil " << std::endl;
+         std::cout << std::endl;
+         std::cout << "x: number of processes in x direction" << std::endl;
+         std::cout << "y: number of processes in y direction" << std::endl;
+         std::cout << "z: number of processes in z direction" << std::endl;
+         std::cout << "px: periodic in x direction?" << std::endl;
+         std::cout << "py: periodic in y direction?" << std::endl;
+         std::cout << "pz: periodic in z direction?" << std::endl;
+         std::cout << "iterations: number of communications per case" << std::endl;
+         std::cout << "messageSize: size of the SendBuffer in bytes" << std::endl;
+         std::cout << "stencil: communication stencil (D3Q27, D3Q19, D3Q7)" << std::endl;
+
+      }
+      return EXIT_FAILURE;
+   }
+
+   Vector3<uint_t> procs;
+   procs[0] = std::stoul(argv[1]);
+   procs[1] = std::stoul(argv[2]);
+   procs[2] = std::stoul(argv[3]);
+   WALBERLA_CHECK_EQUAL(procs[0]*procs[1]*procs[2], mpi::MPIManager::instance()->numProcesses());
+   Vector3<bool> periodicity;
+   periodicity[0] = std::stoul(argv[4]);
+   periodicity[1] = std::stoul(argv[5]);
+   periodicity[2] = std::stoul(argv[6]);
+
+   uint_t iterations  = std::stoul(argv[7]);
+   uint_t messageSize = std::stoul(argv[8]);
+   std::string stencil = argv[9];
+   WALBERLA_CHECK_EQUAL(stencil, "D3Q27", "only D3Q27 is supported!");
+
+   WALBERLA_LOG_DEVEL_VAR_ON_ROOT(procs);
+   WALBERLA_LOG_DEVEL_VAR_ON_ROOT(periodicity);
+   WALBERLA_LOG_DEVEL_VAR_ON_ROOT(iterations);
+   WALBERLA_LOG_DEVEL_VAR_ON_ROOT(messageSize);
+   WALBERLA_LOG_DEVEL_VAR_ON_ROOT(stencil);
+
+   MPIInfo mpiInfo(procs, periodicity);
+
+   WcTimingPool tp;
+   WALBERLA_MPI_BARRIER();
+   if (stencil == "D3Q27")
+   {
+      communicate<stencil::D3Q27>(mpiInfo, iterations, messageSize, false, tp);
+      communicate<stencil::D3Q27>(mpiInfo, iterations, messageSize, true, tp);
+   } else if (stencil == "D3Q19")
+   {
+      communicate<stencil::D3Q19>(mpiInfo, iterations, messageSize, false, tp);
+      communicate<stencil::D3Q19>(mpiInfo, iterations, messageSize, true, tp);
+   } else if (stencil == "D3Q7")
+   {
+      communicate<stencil::D3Q7>(mpiInfo, iterations, messageSize, false, tp);
+      communicate<stencil::D3Q7>(mpiInfo, iterations, messageSize, true, tp);
+   } else
+   {
+      WALBERLA_ABORT("stencil not supported: " << stencil);
+   }
+   WALBERLA_LOG_INFO_ON_ROOT(tp);
+
+   WALBERLA_ROOT_SECTION()
+   {
+      std::map< std::string, walberla::int64_t > integerProperties;
+      std::map< std::string, double >            realProperties;
+      std::map< std::string, std::string >       stringProperties;
+
+      integerProperties["procs_x"]      = int64_c(procs[0]);
+      integerProperties["procs_y"]      = int64_c(procs[1]);
+      integerProperties["procs_z"]      = int64_c(procs[2]);
+      integerProperties["priodicity_x"] = int64_c(periodicity[0]);
+      integerProperties["priodicity_y"] = int64_c(periodicity[1]);
+      integerProperties["priodicity_z"] = int64_c(periodicity[2]);
+      integerProperties["iterations"]   = int64_c(iterations);
+      integerProperties["messageSize"]  = int64_c(messageSize);
+      stringProperties["stencil"]       = stencil;
+
+      auto runId = postprocessing::storeRunInSqliteDB( "ProbeVsTwoMessages.sqlite", integerProperties, stringProperties, realProperties );
+      postprocessing::storeTimingPoolInSqliteDB( "ProbeVsTwoMessages.sqlite", runId, tp, "Timings" );
+   }
+
+   return 0;
+}
+} // namespace walberla
+
+int main( int argc, char* argv[] )
+{
+   return walberla::main( argc, argv );
+}
diff --git a/tests/core/CMakeLists.txt b/tests/core/CMakeLists.txt
index bfecc04b9d9ed646224e574bff7db7a87a4830fa..4d3ee6bc70016ab99e0d51f1cd05042974231ed6 100644
--- a/tests/core/CMakeLists.txt
+++ b/tests/core/CMakeLists.txt
@@ -140,9 +140,6 @@ waLBerla_execute_test( NAME SetReductionTest27 COMMAND $<TARGET_FILE:SetReductio
 
 
 
-waLBerla_compile_test( FILES mpi/ProbeVsExtraMessage.cpp DEPENDS postprocessing)
-
-
 ##############
 # selectable #
 ##############
diff --git a/tests/core/mpi/ProbeVsExtraMessage.cpp b/tests/core/mpi/ProbeVsExtraMessage.cpp
deleted file mode 100644
index a6d3377437c4df849bac98d4a3631065dd355d12..0000000000000000000000000000000000000000
--- a/tests/core/mpi/ProbeVsExtraMessage.cpp
+++ /dev/null
@@ -1,216 +0,0 @@
-//======================================================================================================================
-//
-//  This file is part of waLBerla. waLBerla is free software: you can
-//  redistribute it and/or modify it under the terms of the GNU General Public
-//  License as published by the Free Software Foundation, either version 3 of
-//  the License, or (at your option) any later version.
-//
-//  waLBerla is distributed in the hope that it will be useful, but WITHOUT
-//  ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
-//  FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
-//  for more details.
-//
-//  You should have received a copy of the GNU General Public License along
-//  with waLBerla (see COPYING.txt). If not, see <http://www.gnu.org/licenses/>.
-//
-//! \file ProbeVsExtraMessage.h
-//! \author Martin Bauer <martin.bauer@fau.de>
-//! \brief Micro Benchmark, measuring time for different variable sized communications
-//
-//======================================================================================================================
-
-
-#include "core/debug/TestSubsystem.h"
-#include "core/Environment.h"
-#include "core/mpi/MPIManager.h"
-#include "core/math/Random.h"
-#include "core/timing/TimingPool.h"
-#include "postprocessing/sqlite/SQLite.h"
-
-#include <iostream>
-#include <sstream>
-
-namespace walberla {
-
-int getProcToReceiveFrom()
-{
-   auto mpi = MPIManager::instance();
-   int res = mpi->worldRank() -1;
-   if ( res < 0 )
-      res = mpi->numProcesses() -1;
-   return res;
-}
-
-int getProcToSendTo()
-{
-   auto mpi = MPIManager::instance();
-   int res = mpi->worldRank() +1;
-   if ( res == mpi->numProcesses() )
-      res = 0;
-   return res;
-}
-
-int getRandomMessageSize( uint_t maxMessageSize )
-{
-   return int_c( math::intRandom( maxMessageSize / 3, maxMessageSize-1 ) );
-}
-
-void iprobeVersion( uint_t iterations, uint_t maxMessageSize, WcTimingPool & timingPool )
-{
-   char * sendBuf = new char[ maxMessageSize ];
-   char * recvBuf = new char[ maxMessageSize ];
-
-   auto & timer = timingPool["iprobe"];
-
-   for( uint_t i =0; i < iterations; ++i )
-   {
-      int messageSize = getRandomMessageSize( maxMessageSize );
-
-      timer.start();
-      MPI_Request sendRequest;
-      MPI_Isend( sendBuf, messageSize, MPI_BYTE, getProcToSendTo(), 0, MPI_COMM_WORLD, &sendRequest );
-
-      MPI_Status probeStatus;
-      MPI_Probe( getProcToReceiveFrom(), 0, MPI_COMM_WORLD, &probeStatus );
-
-      int count = 0;
-      MPI_Get_count( &probeStatus, MPI_BYTE, &count );
-      MPI_Recv( recvBuf, count, MPI_BYTE, getProcToReceiveFrom(), 0, MPI_COMM_WORLD, MPI_STATUS_IGNORE );
-      MPI_Waitall( 1, &sendRequest, MPI_STATUSES_IGNORE );
-      timer.end();
-   }
-
-   delete [] sendBuf;
-   delete [] recvBuf;
-}
-
-
-void twoMessageVersion( uint_t iterations, uint_t maxMessageSize, WcTimingPool & timingPool )
-{
-   char * sendBuf = new char[ maxMessageSize ];
-   char * recvBuf = new char[ maxMessageSize ];
-
-   auto & timer = timingPool["twoMessages"];
-
-   int recvSize;
-   MPI_Request sendRequests[2];
-
-   const int TAG_SIZE_MSG    = 1;
-   const int TAG_CONTENT_MSG = 0;
-
-   for( uint_t i =0; i < iterations; ++i )
-   {
-      int sendSize = getRandomMessageSize( maxMessageSize );
-
-      timer.start();
-
-      MPI_Request recvSizeMsgRequest;
-      MPI_Irecv( &recvSize, 1      , MPI_INT,  getProcToReceiveFrom(), TAG_SIZE_MSG,    MPI_COMM_WORLD, &recvSizeMsgRequest );
-      MPI_Isend( &sendSize, 1      , MPI_INT,  getProcToSendTo(),      TAG_SIZE_MSG,    MPI_COMM_WORLD, &sendRequests[0]    );
-      MPI_Isend( sendBuf , sendSize, MPI_BYTE, getProcToSendTo(),      TAG_CONTENT_MSG, MPI_COMM_WORLD, &sendRequests[1]    );
-
-      // wait for size message to arrive
-      MPI_Waitall( 1, &recvSizeMsgRequest, MPI_STATUSES_IGNORE );
-      // receive content
-      MPI_Recv( recvBuf, recvSize, MPI_BYTE, getProcToReceiveFrom(), TAG_CONTENT_MSG, MPI_COMM_WORLD, MPI_STATUS_IGNORE );
-
-      // wait for sends to finish
-      MPI_Waitall( 2, sendRequests, MPI_STATUSES_IGNORE );
-
-      timer.end();
-   }
-
-   delete [] sendBuf;
-   delete [] recvBuf;
-
-}
-
-void maxMessageSizeVersion( uint_t iterations, uint_t maxMessageSize, WcTimingPool & timingPool )
-{
-   char * sendBuf = new char[ maxMessageSize ];
-   char * recvBuf = new char[ maxMessageSize ];
-
-   auto & timer = timingPool["maxMessageSizeKnown"];
-
-   for( uint_t i =0; i < iterations; ++i )
-   {
-      int messageSize = getRandomMessageSize( maxMessageSize );
-      timer.start();
-      MPI_Request sendRequest;
-      MPI_Request recvRequest;
-      MPI_Irecv( recvBuf, int_c(maxMessageSize), MPI_BYTE,  getProcToReceiveFrom(), 0, MPI_COMM_WORLD, &recvRequest );
-      MPI_Isend( sendBuf, messageSize,           MPI_BYTE,  getProcToSendTo(),      0, MPI_COMM_WORLD, &sendRequest );
-
-      MPI_Status status;
-      MPI_Waitall( 1, &recvRequest, &status );
-
-      int count = 0;
-      MPI_Get_count( &status, MPI_BYTE, &count );
-      MPI_Waitall( 1, &sendRequest, MPI_STATUSES_IGNORE );
-      timer.end();
-   }
-
-   delete [] sendBuf;
-   delete [] recvBuf;
-}
-
-
-
-
-
-int main( int argc, char ** argv )
-{
-   debug::enterTestMode();
-   mpi::Environment mpiEnv( argc, argv );
-   MPIManager::instance()->useWorldComm();
-
-   using namespace std;
-
-   WALBERLA_ROOT_SECTION() {
-      if ( argc != 3 && argc != 4 ) {
-         cerr << "Wrong number of arguments " << endl;
-         cerr << "Usage ./probeVsExtraMessage <iterations> <messageSize> " << endl;
-      }
-   }
-   uint_t iterations;
-   uint_t messageSize;
-   std::string arg1( argv[1] );
-   std::string arg2( argv[2] );
-   std::stringstream streamIterations  ( arg1 );
-   std::stringstream streamMessageSize ( arg2 );
-   streamIterations >> iterations;
-   streamMessageSize >> messageSize;
-
-   WcTimingPool tp;
-   iprobeVersion         ( iterations, messageSize, tp );
-   twoMessageVersion     ( iterations, messageSize, tp );
-   maxMessageSizeVersion ( iterations, messageSize, tp );
-
-
-
-   WALBERLA_ROOT_SECTION() {
-      cout << tp << endl;
-   }
-   if( argc == 4) {
-
-      const auto reducedTimeloopTiming = tp.getReduced();
-      WALBERLA_ROOT_SECTION() {
-         std::string dbFile( argv[3] );
-         std::map<std::string, walberla::int64_t> integerProperties;
-         integerProperties["iterations"] = int64_c(iterations);
-         integerProperties["messageSize"] = int64_c(messageSize);
-         integerProperties["processes"] = int64_c(mpi::MPIManager::instance()->numProcesses());
-         postprocessing::SQLiteDB db( dbFile );
-         auto runid = db.storeRun( integerProperties, std::map<std::string, std::string>(), std::map<string, double>());
-         db.storeTimingPool( runid, *reducedTimeloopTiming, "timings" );
-      }
-   }
-
-   return 0;
-}
-} // namespace walberla
-
-int main( int argc, char* argv[] )
-{
-  return walberla::main( argc, argv );
-}
\ No newline at end of file