Shared Research Computing Guidelines

In order to insure that users at the SSDS can complete their research, please adhere to the following guidelines.

  • Users should keep in mind that these servers are shared resources, and their activities may affect other users. The more jobs you execute, the fewer resources are available to others on the system. Whenever possible, please run your jobs sequentially instead of simultaneously. In most cases, users should not be running more than one job at a time.
  • Stata memory management is such that users must have contiguous memory space to execute their jobs (unlike other statistical and mathematical applications). Stata users therefore should not use more memory than necessary. In the past, we have not placed limits on Stata memory allocation because many users need large amounts for large data problems. If Stata users minimize the number of jobs they run and the amount of memory they allocate, we can continue with this policy.
  • Under UNIX, jobs that take more than an hour to run should be niced to make sure they share resources well with shorter jobs. If you prefix your commands with "nice" (e.g. "nice nohup stata test.prg & "), your job will not use a disproportionate amount of resources.
  • Large data sets should be compressed when not in use. The UNIX "gzip" command can be used to compress data in order to maximize disk space. Disk space has not been a problem, but as usage increases so will demand for space. Large inactive files can be archived to DVDs by the SSDS upon request.  Under Windows, files should only temporarily be loaded on local disk space, and the U: home drive should be used for long term storage.

If you need help or recommendations concerning these guidelines, please contact John Daniels, the SSDS statistical programming consultant.