The TACC Global Shared File System, Stockyard, is mounted on nearly all TACC HPC resources as the /work ($WORK) directory. This file system is accessible to all TACC users, and therefore experiences a huge amount of I/O activity (reading and writing to disk) as users run their jobs. This document presents best practices for reducing and mitigating such activity to keep all systems running at maximum efficiency for all TACC users.

What is I/O?

I/O stands for Input/Output and refers to the idea that for every input to a computer (keyboard input, mouse click, external disk access), there is an output (to the screen, in game play, write to disk). In the HPC environment, I/O refers almost exclusively to disk access: opening and closing files, reading from, writing to, and searching within files. Each of these I/O operations (iops), "open", "write", and "close", access each file system's MetaData Server (MDS). The MDS coordinates access to the /work file system for all users. If a file system's MDS is overwhelmed by a user's I/O workflow activities, then that file system could go down for an indeterminate period and all current jobs on that resource may fail.

Examples of intensive I/O activity that could affect the system include, but are not limited to:

• reading/writing 100+ GBs to checkpoint or output files
• running with 4096+ MPI tasks all reading/writing individual files
• Hundreds of concurrent Python jobs, especially those using multiple modules such as pandas, numpy, matplotlib, mpi4py, etc.

As TACC's user base continues to expand, the stress on the resources' shared file systems increases daily. TACC staff now recommends new file system and job submission guidelines in order to maintain file system stability. If a user's jobs or activities are stressing the file system, then every other user's jobs and activities are impacted, and the system admins may resort to cancelling the user's jobs and suspending access to the queues.

If you know your jobs will generate significant I/O, please submit a support ticket and an HPC consultant will work with you.

Recommended File Systems Usage

Consider that your /home ($HOME) and /work ($WORK) directories are for storage and keeping track of important items. The $WORK file system is intended to be an area where you can build your code, store your input and output data, and any intermediate results. The $WORK fileystem is not designed to handle jobs with large amounts of I/O load or iops.

Actual job activity, reading and writing to disk, should be offloaded to your resource's $SCRATCH file system. You can start a job from anywhere but the actual work of the job should occur only on the $SCRATCH partition. You can save original items to $HOME or $WORK so that you can copy them over to $SCRATCH if you need to re-generate results.

Table 1 outlines TACC's new recommended guidelines for file system usage. Note that two TACC systems, Longhorn and Maverick2, differ in their file system configurations. Longhorn does not mount the Stockyard file system and therefore has no $WORK access. Maverick2 has no /scratch file system.

Table 1. TACC File System Usage Recommendations

1 Longhorn is not connected to Stockyard so has no /work mount. Consult the Longhorn User Guide's File Systems section for further guidance.

2 Maverick2 does not have its own $SCRATCH file system. Consult the Maverick2 User Guide's File Systems section for further guidance.

3 The operating system updates a file's access time when that file is modified on a login or compute node. Reading or executing a file/script on a login node does not update the access time, but reading or executing on a compute node does update the access time. This approach helps us distinguish between routine management tasks (e.g. tar, scp) and production use. Use the command "ls -ul" to view access times.

Best Practices for Minimizing I/O

Here we present guidelines aimed at minimizing I/O impact on all TACC resources. Primarily, this means redirecting I/O activity away from Stockyard (the $WORK file system) onto each resource's own local storage: usually the respective /tmp or $SCRATCH file systems.

Use Each Compute Node's /tmp Storage Space

Your jobs are run on the compute nodes of each resource and each compute node has a local /tmp directory on it. You can use the /tmp partition to read/write temporary files that do not need to be accessed by other tasks. If this output data is needed at the end of the job, the files may be copied from /tmp to your $SCRATCH directory at the end of your batch script. This will greatly reduce the load on the file system and may provide performance improvement.

Data stored in the /tmp directory is as temporary as its name indicates, lasting only for the duration of your job. Each MPI task will write output to the /tmp directory on the node on which it is running. MPI tasks cannot access data from /tmp on different nodes. Each TACC resource's compute nodes host a different amount of /tmp space as shown in Table 2 below. Submit a support ticket for more help using this directory/storage.

Table 2. TACC Resources Compute Node (/tmp) Storage

Run Jobs Out of Each Resource's Scratch File System

Each TACC resource (except Maverick2) has its own Scratch file system, /scratch, accessible by the $SCRATCH environment variable and the cds alias.

Scratch file systems are not shared across TACC production systems but are specific to one resource. Scratch file systems have neither file count or file size quotas, but are subject to periodic and unscheduled file purges should total disk usage exceed a safety threshold.

TACC staff recommends you run your jobs out of your resource's $SCRATCH file system instead of the global $WORK file system. To run your jobs out of $SCRATCH, copy (stage) the entire executable/package along with all needed job input files and/or needed libraries to your resource's $SCRATCH directory.

Compute nodes should not reference the $WORK file system unless it's to stage data in or out, and only before or after jobs.

Your job script should also direct the job's output to the local scratch directory:

# stage executable and data
cd $SCRATCH
mkdir testrunA
cp $WORK/myprogram testrunA
cp $WORK/jobinputdata testrunA

# launch program
ibrun testrunA/myprogram testrunA/myinputdata > testrunA/output

# copy results back permanent storage once job is done
cp testrunA/output $WORK/savetestrunA

Avoid Writing One File Per Process

If your program regularly writes data to disk from each process, for instance for checkpointing, avoid writing output to a separate file for each process, as this will quickly overwhelm the Metadata Server. Instead, employ a library such as hdf5 or netcdf to write a single parallel file for the checkpoint. A one-time generation of one file per process (for instance at the end of your run) is less serious, but even then you should consider writing parallel files.

Alternatively, you could write these per-process files to each compute node's /tmp directory, see below.

Avoid Repeated Reading/Writing to the Same File

Jobs that have multiple tasks that read and/or write to the same file will often suspend the file in question in an open state in order to accommodate the changes happening to it. Please make sure that your I/O activity is not being directed to a single file repeatedly. You can use /tmp on the node to store this file if the condition cannot be avoided. If you require shared file operations, then please ensure your I/O is optimized.

If you anticipate the need for multiple nodes or processes to write to a single file in parallel (aka single file with multiple writers/collective writers), please submit a support ticket for assistance.

Monitor Your File System Quotas

If you are close to file quota on either the $WORK or $HOME file system, your job may fail due to being unable to write output, and this will cause stress to the file systems when attempting to write beyond quota. It's important to monitor your disk and file usage on all TACC resources where you have an allocation.

Monitor your file system's quotas and usage using the taccinfo command. This output displays whenever you log on to a TACC resource.

---------------------- Project balances for user  ----------------------
| Name           Avail SUs     Expires | Name           Avail SUs     Expires |
| Allocation             1             | Alloc                100             |
------------------------ Disk quotas for user   -------------------------
| Disk         Usage (GB)     Limit    %Used   File Usage       Limit   %Used |
| /home1              1.5      25.0     6.02          741      400000    0.19 |
| /work             107.5    1024.0    10.50         2434     3000000    0.08 |
| /scratch1           0.0       0.0     0.00            3           0    0.00 |
| /scratch3       41829.5       0.0     0.00       246295           0    0.00 |
-------------------------------------------------------------------------------

Manipulate Data in Memory, not on Disk

Manipulate data in memory instead of files on disk when necessary. This means:

• For unimportant data that do not need a backup, process that data directly in memory.
• For any commands in the intermediate steps, process those commands directly in memory instead of creating extra script files for them.

Stripe Large Files on $SCRATCH and $WORK

When transferring or creating large files, it's important that you stripe the receiving directory. See the respective "Striping Large Files" sections in the Stampede2 and Frontera user guides.

Govern I/O with OOOPS

TACC staff has developed OOOPS, Optimal Overloaded I/O Protection System, an easy-to-use tool to help HPC users optimize heavy I/O requests and reduce the impact of high I/O jobs. If your jobs have a particularly high I/O footprint, then you must employ the OOOPS tool to govern that I/O activity.

Employing OOOPS may slow down your job significantly if your job has a lot of I/O.

The OOOPS module is currently installed on TACC's Frontera and Stampede2 resources.

Functions

The OOOPS module provides two functions "set_io_param" and "set_io_param_batch" for single-node jobs and multiple-node jobs, respectively. These commands adjust the maximum allowed frequency of "open" and "stat" function calls on all compute nodes involved in a running job. Execute these two commands within a Slurm job script or within an idev session.

These functions instruct the system to modulate your job's I/O activity, thus reducing the impact on the designated file system. For both functions, use "0" to indicate the $SCRATCH file system and "1" to indicate the $WORK file system. Note: these indices are subject to change. See each command's help option to ensure correct parameters:

c123-456$ set_io_param -h

Indicate the frequency of "open" and "stat" function calls, from the least to the most, with "low", "medium", or "high".

How to Use OOOPS

First, load the "ooops" module in your job script or idev session to deploy OOOPS. Next, set the frequency of I/O activities using either the "set_io_param" or "set_io_param_batch" command.

Example: Single-Node Job on $SCRATCH

Example: Multi-Node Job on $SCRATCH

I/O Warning

If OOOPS finds intensive I/O work in your job, it will print out warning messages and create an open/stat call report after the job finishes. To enable reporting, load the OOOPS module on a login node, and then submit your batch job. The reporting function will not be enabled if the module is loaded within a batch script.

Developers

Contact the OOOPS developers, Lei Huang and Si Liu, with any OOOPS related questions.

Python I/O Management

For jobs that make use of large numbers of Python modules or use local installations of Python/Anaconda/MiniConda, TACC staff provides additional tools to help manage the I/O activity caused by library and module calls.

On Stampede2 and Frontera: Load the "python_cacher" module in your job script:

module load python_cacher

This library will cache python modules to local disk so python programs won't keep pulling the modules over and over from the "/scratch" or "/work" file systems.

In case python_cacher does not work, you can copy your Python/Anaconda/MiniConda directory to the local /tmp directory of each involved compute node for a specific job.

Tracking Job I/O

Stampede2 and Frontera: To track the full extent of your I/O activity over the course of your job, you can employ another TACC tool, "iomonitor" that will report on "open()" and "stat()" calls during your job's run. Place the following lines in your job submission script after your Slurm commands, to wrap your executable:

export LD_PRELOAD=/home1/apps/tacc-patches/io_monitor/io_monitor.so:\
    /home1/apps/tacc-patches/io_monitor/hook.so
ibrun my_executable
unset LD_PRELOAD

Log files will be generated during the job run in the working directory with prefix "log_io_*."

Note: Since the iomonitor tool may itself generate a lot of files, we highly recommend you profile your job beginning with trivial cases, then ramping up to the desired number of nodes/tasks.

Please feel free to submit a support ticket if you need any further assistance.