OSG Connect provides two locations for uploading files (data and software) that are needed for running jobs:
home: /home/<username> (general storage from which ALL jobs should be submitted) public: /public/<username> (only for large job input and output using stash links (input) and stashcp (output))
In general, OSG Connect users are responsible for managing data in these folders and for using appropriate mechanisms for delivering data to/from jobs, as detailed below. Each is controlled with a quota and should be treated as temporary storage for active job execution. OSG Connect has no routine backup of data in these locations, and users should remove old data after jobs complete, in part, to make room for future submissions. If you think you'll need more space for a set of concurrently-queued jobs, even after cleaning up old data, please send a request to email@example.com!
For additional data information, see also the "Data Storage and Transfer" section of our FAQ.
Note: OSG Connect staff reserve the right to monitor and/or remove data without notice to the user IF doing so is necessary for ensuring proper use or to quickly fix a performance or security issue.
Data Locations and Quotas
Your OSG Connect account includes access to two data storage locations:
/public. Where you store your files and how your files
are made accessible to your jobs depends on the size of the file and how
much data is needed or produced by your jobs.
|Location||Storage Needs||Network mounted||Backed Up?||Initial Quota|
||Storage of submit files, input files <100MB each, and per-job output
up to a 1GB. Jobs should ONLY be submitted from this folder.
||Staging ONLY for large input files (100MB-50GB, each) for publicly-accessible
download into jobs (using
Your quota status will be displayed when you connect to your OSG Connect login node:
Disk utilization for username: /public : [ ] 0% (0/500000 MB) /home : [ # ] 4% (2147/53687 MB)
You can also display your quota usage at any time using the command
quota while connected to your login node.
Don't hesitate to contact us at firstname.lastname@example.org if you think you need a quota increase! We can support very large amounts of data.
/home Usage And Policies
User directories within
/home are meant for general-use storage of your files
needed for job submission. The initial quota per user is 50 GBs, and can be increased by request
to email@example.com when a user needs more space for appropriately-sized files.
ALL JOBS MUST BE SUBMITTED FROM WITHIN
/home. Users are also prohibited
from making their
/home directory world-readable due to security concerns. See
Policies for Using OSG via OSG Connect Submit Servers for more details.
If you're unable to submit jobs or your jobs are going on hold because you've
/home quota, please contact us at
firstname.lastname@example.org about a quota increase.
/public Usage and Policies
User directories within
/public are meant ONLY for staging job files too large for regular
HTCondor file transfer (per-job input greater than 100MB, or per-job output greater than 1GB), and should only OSG caching
mechanisms (see tables for input and output, further below).
JOBS MUST NEVER BE SUBMITTED FROM WITHIN
/public, and should not list
/public files in the
transfer_input_files or other lines of a submit file, unless as a
stash:/// address (see tables further below).
Files placed in
/public should only be accessed by jobs using the below tools (see
Transferring Data To/From Jobs). Users violating these policies may
lose the ability to submit jobs until their submissions are corrected.
The initial disk quota of
/public is 500 GBs. Contact email@example.com
if you will need an increase for concurrently running work, after cleaning up all
data from past jobs. Given that users should not be storing long-term data (like submit files,
software, etc.) in
/public, files and directories that have not been accessed for over six
months may be deleted by OSG Connect staff with or without notifying the user.
Files placed within a user's
/public directory are publicly accessible,
discoverable and readable by anyone. Data is made public via the
stash transfer mechanisms (which also make data public via http/https), and mirrored
to a shared data repository which is available on a large number of systems around the world.
Is there any support for private data?
If you do not want your data to be downloadable by anyone, and it's small enough for
HTCondor file transfer, then it should be staged in your
/home directory and
transferred to jobs with HTCondor file transfer (
transfer_input_files in the submit
file). If it cannot be public (cannot use http or stash for job delivery), and is too
large for HTCondor file transfer, then it's not a good fit for the open environment of
the Open Science Pool, and another resource will likely be more appropriate. As a
reminder, if the data is not being used for active computing work on OSG Connect, it
should not be stored on OSG Connect systems, and our data storage locations are not
backed up or suitable for project-duration storage.
External Data Transfer to/from OSG Connect Login Nodes
In general, common Unix tools such as
scp, Putty, WinSCP,
gFTP, etc. can be used to upload data from your computer to the OSG Connect login node, or to download files from the OSG Connect login node.
Transferring Data To/From Jobs
Transferring Input Data to Jobs
This table summarizes the options for sending input files from the OSG Connect login node to the execution node where a job is running. This assumes that you have already uploaded these input files from your own computer to your OSG Connect login node.
|Transfer Method||File Sizes||File Location||Command||More Info|
<500 MB total per job
||HTCondor File Transfer|
<10 GB per job
|HTTP||<1GB||non-OSG web server||
|GridFTP||> 10 GB||
Transferring Output Data from Jobs
This table summarizes a job's options for returning output files generated by the job back to the OSG Connect login node.
|Transfer Method||File Sizes||Transfer To||Command||More Info|
|regular HTCondor file transfer||< 1 GB||
||HTCondor default output transfer or
|OSG's transfer tools||>1GB and
< 10 GB
|GridFTP||> 10 GB||
Watch this video from the 2021 OSG Virtual School for more information about Handling Data on OSG:
This page was updated on May 20, 2022 at 18:10 from start/data/osgconnect-storage.md.