The "sastemp" directories are provided for temporary, shared workspace storage, and are not intended for reliable permanent storage of files. Each sastemp directory is 2TB of working space. The output may be saved there at the completion of a program, but only to be IMMEDIATELY removed.
Any file in /sastemp which is either not in use by an active program or being transferred to a user's homeschool can be deleted without prior notice. Files on /sastemp are deleted automatically every 48 hours. Please note: Files kept in your home directory will never be deleted.
- Files located here are not backed up!
- Temporary work libraries created by SAS (i.e., SAS_work* subdirectories) are automatically removed when the associated SAS process has ended.
- If anyone of the /sastemp directories is more than 95% full, the user accounting for the most space on that particular file system will have all of his/her files deleted from it without notice, regardless of file age.
- Beyond the specific criteria listed above, WRDS may delete /sastemp files without notice if they are being used for long-term storage or a single user's files are preventing others from access to adequate Sastemp space. Fortunately, the latter has been a rare event, and there is often an opportunity to contact the owner before file deletion.
As such, /sastemp is generally to be used for temporary storage of data that your running research program is actively using, and never used for any long-term storage. Remember, this directory is shared with all other WRDS users!
Batch jobs additionally have the following CPU and RAM usage restrictions:
- By default, a WRDS Cloud job is assigned to 2 cores (for a total of 4 threads) per job.
- Users may request up to 8 cores (for a total of 16 threads) per job.
- By default, a WRDS Cloud job is assigned 8GB per core (for a total of 16GB if left at the default core count).
- Users may request up to 48GB RAM total per job (calculated as RAM * cores requested).
For more information on requesting additional CPU and RAM resources, please see Batch Jobs on the WRDS Cloud.