Changes between Version 28 and Version 29 of Public/User_Guide/Batch_system
- Timestamp:
- Apr 6, 2020, 8:37:10 AM (4 years ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
Public/User_Guide/Batch_system
v28 v29 228 228 229 229 In order to establish MPI communication across modules using different interconnect technologies, some special Gateway nodes must be used. On the DEEP-EST system, MPI communication across gateways is needed only between Infiniband and Extoll interconnects. 230 230 231 **Attention:** Only !ParaStation MPI supports MPI communication across gateway nodes. 231 232 … … 239 240 # Use the packjob feature to launch separately CM and DAM executable 240 241 241 #SBATCH --job-name=imb 242 243 # General configuration of the job 244 #SBATCH --job-name=modular-imb 242 245 #SBATCH --account=deep 243 #SBATCH --output=IMB-%j.out 244 #SBATCH --error=IMB-%j.err 245 #SBATCH --time=00:05:00 246 #SBATCH --time=00:10:00 247 #SBATCH --output=modular-imb-%j.out 248 #SBATCH --error=modular-imb-%j.err 249 250 # Configure the gateway daemon 246 251 #SBATCH --gw_num=1 247 #SBATCH --gw_binary=/opt/parastation/bin/psgwd.extoll248 252 #SBATCH --gw_psgwd_per_node=1 249 253 254 # Configure node and process count on the CM 250 255 #SBATCH --partition=dp-cn 251 256 #SBATCH --nodes=1 252 #SBATCH --ntasks =1257 #SBATCH --ntasks-per-node=1 253 258 254 259 #SBATCH packjob 255 260 261 # Configure node and process count on the DAM 256 262 #SBATCH --partition=dp-dam-ext 257 263 #SBATCH --nodes=1 258 #SBATCH --ntasks=1 259 264 #SBATCH --ntasks-per-node=1 265 266 # Echo job configuration 260 267 echo "DEBUG: SLURM_JOB_NODELIST=$SLURM_JOB_NODELIST" 261 268 echo "DEBUG: SLURM_NNODES=$SLURM_NNODES" 262 269 echo "DEBUG: SLURM_TASKS_PER_NODE=$SLURM_TASKS_PER_NODE" 263 270 264 # Execute 265 srun hostname : hostname 266 srun module_dp-cn.sh : module_dp-dam-ext.sh 267 }}} 268 269 It uses two execution scripts for loading the correct environment and starting the IMB on the CM and the DAM node (this approach can also be used to start different programs, e.g. one could think of a master and worker use case). The execution scripts could look like: 270 271 {{{ 272 #!/bin/bash 273 # Script for the CM using InfiniBand 274 271 272 # Set the environment to use PS-MPI 275 273 module --force purge 276 274 module use $OTHERSTAGES … … 279 277 module load ParaStationMPI 280 278 281 # Execution 282 EXEC=$PWD/mpi-benchmarks/IMB-MPI1 283 ${EXEC} PingPong 284 }}} 285 286 {{{ 287 #!/bin/bash 288 # Script for the DAM using Extoll 289 290 module --force purge 291 module use $OTHERSTAGES 292 module load Stages/Devel-2019a 293 module load Intel 294 module load ParaStationMPI 295 296 # Execution 297 EXEC=$PWD/mpi-benchmarks/IMB-MPI1 298 ${EXEC} PingPong 299 }}} 279 # Show the hosts we are running on 280 srun hostname : hostname 281 282 # Execute 283 APP="/p/project/cfa_partec/pickartz/mpi-benchmarks/src_c/IMB-MPI1 Uniband" 284 srun ${APP} : ${APP} 285 }}} 286 287 300 288 301 289 **Attention:** During the first part of 2020, only the DAM nodes will have Extoll interconnect, while the CM and the ESB nodes will be connected via Infiniband. This will change later during the course of the project (expected Summer 2020), when the ESB will be equipped with Extoll connectivity (Infiniband will be removed from the ESB and left only for the CM).