Skip to content
GitLab

User guide · Changes

Page history
Updated User-guide.md (markdown) authored by Anonymous's avatar Anonymous
Show whitespace changes
Inline Side-by-side
User-guide.md
View page @ bdd28f09
# Job monitoring and optimization with EAR
EAR was first designed to be usable 100% transparently by users, which means that
you can run your applications enabling/disabling/tuning EAR with the less effort
EAR was first designed to be 100% transparent by users, which means that
you can run your applications enabling/disabling/tuning EAR with minimal effort
for changing your workflow, e.g., submission scripts.
This is achieved by providing integrations (e.g., plug-ins, hooks) with system batch
schedulers, which do all the effort to set-up EAR at job submission.
By now, **[SLURM](https://slurm.schedmd.com/documentation.html) is the batch scheduler full compatible with EAR** thanks to EAR's SLURM SPANK plug-in.
By now, **[SLURM](https://slurm.schedmd.com/documentation.html) is the batch scheduler fully compatible with EAR** thanks to EAR's SLURM SPANK plug-in.
With EAR's SLURM plug-in, running an application with EAR is as easy as submitting
a job with either `srun`, `sbatch` or `mpirun`. The EAR Library (EARL) is automatically
......@@ -39,10 +39,10 @@ EAR supports the utilization of both `mpirun`/`mpiexec` and `srun` commands.
When using `sbatch`/`srun` or `salloc`, [Intel MPI](https://www.intel.com/content/www/us/en/developer/tools/oneapi/mpi-library.html#gs.mufipm)
and [OpenMPI](https://www.open-mpi.org/) are fully supported.
When using specific MPI flavour commands to start applications (e.g., `mpirun`,
`mpiexec.hydra`), there are some keypoints which you must take account.
`mpiexec.hydra`), there are some key points which you must take account.
See [next sections](#using-mpirunmpiexec-command) for examples and more details.
Review SLURM's [MPI Users Guide](https://slurm.schedmd.com/mpi_guide.html), read your cluster documentation or ask your system administrator to see how SLURM is integrated with the MPI Library in your system.
Review SLURM's [MPI Users Guide](https://slurm.schedmd.com/mpi_guide.html), read your cluster documetation or ask your system administrator to see how SLURM is integrated with the MPI Library in your system.
### Hybrid MPI + (OpenMP, CUDA, MKL) applications
......@@ -51,10 +51,10 @@ EARL automatically supports this use case.
### Python and Julia MPI applications
EARL cannot detect automatically MPI symbols when some of these languages is used.
EARL cannot detect automatically MPI symbols when some of these languages are used.
On that case, an environment variable is provided to give EARL a hint of the MPI flavour being used.
Export [`EAR_LOAD_MPI_VERSION`](EAR-environment-variables#ear_load_mpi_version) environment with the value from the following table depending on the MPI implementation you are loading:
Export the [`EAR_LOAD_MPI_VERSION`](EAR-environment-variables#ear_load_mpi_version) environment with the value from the following table depending on the MPI implementation you are loading:
| MPI flavour | Value |
| ----------- | ----- |
......@@ -122,21 +122,21 @@ Check its [section](#mpi4py) for running this kind of use case.
### Python
Since version 4.1 EAR automatically executes the Library with Python applications, so no action is needed.
Since version 4.1 EAR automatically executes the Library with Python applications, so no further action is needed.
You must run the application with `srun` command to pass through the EAR's SLURM SPANK plug-in in order to enable/disable/tuning EAR.
See [EAR submission flags](#ear-job-submission-flags) provided by EAR SLURM integration.
### OpenMP, CUDA, Intel MKL and OneAPI
To load EARL automatically with non-MPI applications it is required to have it compiled with dynamic symbols and also it must be executed with `srun` command.
For example, for CUDA applications the `--cudart=shared` option must be used at compile time.
For example, for CUDA applications you must use the `--cudart=shared` option at compile time.
EARL is loaded for OpenMP, MKL and CUDA programming models when symbols are dynamically detected.
## Other application types or frameworks
For other programming models or sequential apps not supported by default, EARL can
be forced to be loaded by setting [`EAR_LOADER_APPLICATION`](EAR-environment-variables#ear_loader_application)
enviroment variable, which must be defined with the executable name.
environment variable, which must be defined with the executable name.
For example:
```
......@@ -150,7 +150,7 @@ srun my_app
[Apptainer](https://apptainer.org/) (formerly Singularity) is an open source technology for containerization.
It is widely used in HPC contexts because the level of virtualization it offers enables the access to local services.
It allows for geater reproducibility, making the programs less dependant on the environment they are being run on.
It allows for greater reproducibility, making the programs less dependent on the environment they are being run on.
An example singularity command could look something like this:
......@@ -171,12 +171,12 @@ To bind folders there are two options: (1) using the environment variable `SINGU
Specifying *path_2* and *perm* is optional.
If they are not specified *path_1* will be bound in the same location.
To make EAR working the following paths sould be added to the binding configuration:
To make EAR working the following paths should be added to the binding configuration:
- `$EAR_INSTALL_PATH,$EAR_INSTALL_PATH/bin,$EAR_INSTALL_PATH/lib,$EAR_TMP`
You should have an EAR module to have the above environment variables.
Contact with your system administrator for more information.
Contact your system administrator for more information.
Once paths are deployed, to execute (for example) an OpenMPI application inside a Singularity/Apptainer enabling the EAR Library just the following is needed:
......@@ -206,7 +206,7 @@ See [next section](#runtime-report-plug-ins) about report plug-ins.
COMP Superscalar ([COMPSs](https://compss-doc.readthedocs.io/en/latest/index.html)) is a task-based programming model which aims to ease the development of applications for distributed infrastructures, such as large High-Performance clusters (HPC), clouds and container managed clusters.
COMPSs provides a programming interface for the development of the applications and a runtime system that exploits the inherent parallelism of applications at execution time.
**Since version 5.0 EAR supports monitoring and optimization of workflows** and the COMPSs Framework includes the integration with EAR.
**Since version 5.0 EAR supports monitoring and optimization of workflows** and the COMPSs Framework includes integration with EAR.
Check out the [dedicated section](https://compss-doc.readthedocs.io/en/latest/Sections/05_Tools/05_EAR.html#) from the official COMPSs documentation for more information about how to measure the energy consumption of your workflows.
EARL loading is **only available** using `enqueue_compss` and with Python applications.
......@@ -225,7 +225,7 @@ As a very simple hint of your application workload, you can enable EARL verbosit
**The information is shown at _stderr_ by default.**
Read how to set up verbosity at [submission time](#ear-job-submission-flags) and
[verbosity environment variables](EAR-environment-variables#verbosity) provided
for a more advanced tunning of this EAR feature.
for a more advanced tuning of this EAR feature.
## Post-mortem application data
......@@ -500,7 +500,7 @@ It has an internal system to avoid repeating functions that are executed just on
time per job or node, like SLURM does with its plugins.
**IMPORTANT NOTE** If you are going to launch `n` applications with `erun` command through a sbatch job, you must set the environment variable `SLURM_STEP_ID` to values from `0` to `n-1` before each `mpirun` call.
By this way `erun` will inform the EARD the correct step ID to be stored then to the Database.
By this way `erun` will inform the EARD the correct step ID to be then stored to the Database.
# EAR job Accounting (`eacct`)
......@@ -582,4 +582,4 @@ The Library deals with job monitoring and is the component which implements and
optimization policies based on monitored workload.
**We highly recommend you** to read [EARL](https://gitlab.bsc.es/ear_team/ear/-/wikis/Architecture#the-ear-library-job-manager) documentation and also how energy policies work
in order to better understand what is doing the Library internally, so you will can explore easily all features (e.g., tunning variables, collecting data) EAR offers to the end-user so you will have more knowledge about how much resources your application consumes and how to correlate with its computational characteristics.
in order to better understand what is doing the Library internally, so you can explore easily all features (e.g., tunning variables, collecting data) EAR offers to the end-user so you will have more knowledge about how much resources your application consumes and how to correlate with its computational characteristics.