Large restart file output error in a high resolution configuration for CAM5-EUL dynamic core

E

eaton

CSEG and Liaisons
How much memory and how many cores on a shared memory node?How many tasks are you using per node?  What is the total number of nodes?The output log will contain information about the pio_stride and pio_numtasks.  What are those numbers?  
 
E

eaton

CSEG and Liaisons
How much memory and how many cores on a shared memory node?How many tasks are you using per node?  What is the total number of nodes?The output log will contain information about the pio_stride and pio_numtasks.  What are those numbers?  
 
E

eaton

CSEG and Liaisons
How much memory and how many cores on a shared memory node?How many tasks are you using per node?  What is the total number of nodes?The output log will contain information about the pio_stride and pio_numtasks.  What are those numbers?  
 
E

eaton

CSEG and Liaisons
How much memory and how many cores on a shared memory node?How many tasks are you using per node?  What is the total number of nodes?The output log will contain information about the pio_stride and pio_numtasks.  What are those numbers?  
 
E

eaton

CSEG and Liaisons
How much memory and how many cores on a shared memory node?How many tasks are you using per node?  What is the total number of nodes?The output log will contain information about the pio_stride and pio_numtasks.  What are those numbers?  
 
Z

zhangyi@lasg_iap_ac_cn

Member
Hi eaton, the machine is 12 cpus per node, but I don't know the memory.Normally, I use 400 cpus for this resolution, because this is the maximum cores that EUL-core can use.34 nodes are used, so tasks per node is 12 or less. In this setting, pio_stride=4 pio_numtasks= 100To double the nodes, actually I use 1000 cpus on 84 nodes, of course, only 400 cores for dynamics.pio_stride=4 pio_numtasks= 250.
 
Z

zhangyi@lasg_iap_ac_cn

Member
Hi eaton, the machine is 12 cpus per node, but I don't know the memory.Normally, I use 400 cpus for this resolution, because this is the maximum cores that EUL-core can use.34 nodes are used, so tasks per node is 12 or less. In this setting, pio_stride=4 pio_numtasks= 100To double the nodes, actually I use 1000 cpus on 84 nodes, of course, only 400 cores for dynamics.pio_stride=4 pio_numtasks= 250.
 
Z

zhangyi@lasg_iap_ac_cn

Member
Hi eaton, the machine is 12 cpus per node, but I don't know the memory.Normally, I use 400 cpus for this resolution, because this is the maximum cores that EUL-core can use.34 nodes are used, so tasks per node is 12 or less. In this setting, pio_stride=4 pio_numtasks= 100To double the nodes, actually I use 1000 cpus on 84 nodes, of course, only 400 cores for dynamics.pio_stride=4 pio_numtasks= 250.
 
Z

zhangyi@lasg_iap_ac_cn

Member
Hi eaton, the machine is 12 cpus per node, but I don't know the memory.Normally, I use 400 cpus for this resolution, because this is the maximum cores that EUL-core can use.34 nodes are used, so tasks per node is 12 or less. In this setting, pio_stride=4 pio_numtasks= 100To double the nodes, actually I use 1000 cpus on 84 nodes, of course, only 400 cores for dynamics.pio_stride=4 pio_numtasks= 250.
 
Z

zhangyi@lasg_iap_ac_cn

Member
Hi eaton, the machine is 12 cpus per node, but I don't know the memory.Normally, I use 400 cpus for this resolution, because this is the maximum cores that EUL-core can use.34 nodes are used, so tasks per node is 12 or less. In this setting, pio_stride=4 pio_numtasks= 100To double the nodes, actually I use 1000 cpus on 84 nodes, of course, only 400 cores for dynamics.pio_stride=4 pio_numtasks= 250.
 
Z

zhangyi@lasg_iap_ac_cn

Member
Hi eaton, the machine is 12 cpus per node, but I don't know the memory.Normally, I use 400 cpus for this resolution, because this is the maximum cores that EUL-core can use.34 nodes are used, so tasks per node is 12 or less. In this setting, pio_stride=4 pio_numtasks= 100To double the nodes, actually I use 1000 cpus on 84 nodes, of course, only 400 cores for dynamics.pio_stride=4 pio_numtasks= 250.
 
Z

zhangyi@lasg_iap_ac_cn

Member
Hi eaton, the machine is 12 cpus per node, but I don't know the memory.Normally, I use 400 cpus for this resolution, because this is the maximum cores that EUL-core can use.34 nodes are used, so tasks per node is 12 or less. In this setting, pio_stride=4 pio_numtasks= 100To double the nodes, actually I use 1000 cpus on 84 nodes, of course, only 400 cores for dynamics.pio_stride=4 pio_numtasks= 250.
 
Z

zhangyi@lasg_iap_ac_cn

Member
Hi eaton, the machine is 12 cpus per node, but I don't know the memory.Normally, I use 400 cpus for this resolution, because this is the maximum cores that EUL-core can use.34 nodes are used, so tasks per node is 12 or less. In this setting, pio_stride=4 pio_numtasks= 100To double the nodes, actually I use 1000 cpus on 84 nodes, of course, only 400 cores for dynamics.pio_stride=4 pio_numtasks= 250.
 
E

eaton

CSEG and Liaisons
I would try to double the nodes by requesting 68 nodes and only assign 6 tasks per node.  This gives each task twice as much memory as in the successful configuration used for the 30 level grid.  I would also try setting the namelist variable atm_pio_stride=6.  This will put just 1 pio task on each node which should minimize the overhead incurred when writing the restart file.
 
E

eaton

CSEG and Liaisons
I would try to double the nodes by requesting 68 nodes and only assign 6 tasks per node.  This gives each task twice as much memory as in the successful configuration used for the 30 level grid.  I would also try setting the namelist variable atm_pio_stride=6.  This will put just 1 pio task on each node which should minimize the overhead incurred when writing the restart file.
 
E

eaton

CSEG and Liaisons
I would try to double the nodes by requesting 68 nodes and only assign 6 tasks per node.  This gives each task twice as much memory as in the successful configuration used for the 30 level grid.  I would also try setting the namelist variable atm_pio_stride=6.  This will put just 1 pio task on each node which should minimize the overhead incurred when writing the restart file.
 
E

eaton

CSEG and Liaisons
I would try to double the nodes by requesting 68 nodes and only assign 6 tasks per node.  This gives each task twice as much memory as in the successful configuration used for the 30 level grid.  I would also try setting the namelist variable atm_pio_stride=6.  This will put just 1 pio task on each node which should minimize the overhead incurred when writing the restart file.
 
E

eaton

CSEG and Liaisons
I would try to double the nodes by requesting 68 nodes and only assign 6 tasks per node.  This gives each task twice as much memory as in the successful configuration used for the 30 level grid.  I would also try setting the namelist variable atm_pio_stride=6.  This will put just 1 pio task on each node which should minimize the overhead incurred when writing the restart file.
 
E

eaton

CSEG and Liaisons
I would try to double the nodes by requesting 68 nodes and only assign 6 tasks per node.  This gives each task twice as much memory as in the successful configuration used for the 30 level grid.  I would also try setting the namelist variable atm_pio_stride=6.  This will put just 1 pio task on each node which should minimize the overhead incurred when writing the restart file.
 
E

eaton

CSEG and Liaisons
I would try to double the nodes by requesting 68 nodes and only assign 6 tasks per node.  This gives each task twice as much memory as in the successful configuration used for the 30 level grid.  I would also try setting the namelist variable atm_pio_stride=6.  This will put just 1 pio task on each node which should minimize the overhead incurred when writing the restart file.
 
You must log in or register to reply here.
Back
Top