Hi,
We are running SAS 9.4, and we have IBM Platform Process Manager 8.2 & LSF.
We Wonder what the max number of flows we can run at the same time?
Regards
Morten Fredrik Amundsen
I read this question differently from the other reply.
Instead of thinking how many are possible, I think of how many I want to allow to run at the same time.
The file "lsb.hosts", in <LSF_TOP>/conf/lsbatch/<cluster name>/configdir/lsb.hosts, has controls that you can set to limit (or enable) more jobs to run simultaneously.
The comments at the top of this file explain the settings in more detail and if nothing else will help point you in the correct direction for documentation.
The default is one "job" per CPU. From the lsb.hosts file:
# MXJ is the maximum number of jobs which can run on the host at one time.
# If MXJ is set as ! the system automatically assigns it to be the
# number of CPUs on the host.
A "job" being one SAS job within a SAS flow. SAS flows can have multiple jobs kick off at the same time.
In our case, based on the resources available and how they are used (per the other response), we increased this value past the default to allow more SAS jobs to run at once.
There is no max number.
Memory and CPU constraints will create a practical limit, but the real throttle is on the number of jobs that are running and that can be handled by LSF.
Hi Doug
Thank you for this answer.
I understand by your answer that it's not the number of concurrent flows, but the number of concurrent jobs that may be the problem.
Regards
Morten Fredrik
I read this question differently from the other reply.
Instead of thinking how many are possible, I think of how many I want to allow to run at the same time.
The file "lsb.hosts", in <LSF_TOP>/conf/lsbatch/<cluster name>/configdir/lsb.hosts, has controls that you can set to limit (or enable) more jobs to run simultaneously.
The comments at the top of this file explain the settings in more detail and if nothing else will help point you in the correct direction for documentation.
The default is one "job" per CPU. From the lsb.hosts file:
# MXJ is the maximum number of jobs which can run on the host at one time.
# If MXJ is set as ! the system automatically assigns it to be the
# number of CPUs on the host.
A "job" being one SAS job within a SAS flow. SAS flows can have multiple jobs kick off at the same time.
In our case, based on the resources available and how they are used (per the other response), we increased this value past the default to allow more SAS jobs to run at once.
Hi,
Thank you for the answer. This makes sence. We will look into the file you mention.
Ragards
Morten Fredrik
The real limit is imposed by your main system resources memory, cpu and IO. Whenever you run out of one your system is overtasked. When that happens depends not on LSF but on the usage characteristics of whatever is running on your system and how you tune things. You will have to measure and learn in order to determine ow many concurrent jobs your system can run.
I have learned that no two systems are alike in this respect. I am always pleased with how LSF allows one to adapt easily.
In our case we have a mix of batch jobs and workspace servers. Workspace server (for SAS EG, Studio, Enterprise Miner, ...) are the big unknowns. Some people go wild all day, others start a session, run a small query and go into meetings for the rest of the day without signing off. But also the batch jobs can range from steamrolling our servers to sitting idle for hours waiting for Teradata to complete some big in-database voodoo. So we needed months to figure out what happens and where the limits are. In fact we are now resting at a 1:6 ratio between #CPU and the number of job slots on each host in the grid.
So the best advice I can give is measure, learn, experiment and adjust. No system is the same.
Hope this helps,
-- Jan.
The SAS Users Group for Administrators (SUGA) is open to all SAS administrators and architects who install, update, manage or maintain a SAS deployment.
SAS technical trainer Erin Winters shows you how to explore assets, create new data discovery agents, schedule data discovery agents, and much more.
Find more tutorials on the SAS Users YouTube channel.