GDMS needs to be able to store information in temporary files to operate correctly. The information currently stored in temporary files are the graphs used in the user interface, and active datasets. This is done to avoid the graphs having to be regenerated each time they are drawn. This technique also serves to avoid having images cached in memory, which could result in a large amount of memory being used for data that is not frequently accessed. The GDMS determines when the files in the temporary graph cache are out of date. When this occurs, the graphs are regenerated as required. Temporary files are stored by default in /tmp. The files have randomly generated names which match the regular expression cep.*$. Examples of valid temporary filenames are: cep.5tgSBw cep.7v2DRn cep.AxSGaC cep.d4j1at cep.dinwwj cep.eu4OqI cep.HDxyXc cep.I6U6DO cep.k2Wweg cep.pDpvqC cep.qcRKds cep.sBiOhb cep.VqHGbW cep.YoTRdk cep.zbOzJ3 These files can amount to a non-trivial amount of space on disk. The example files totalled 808 kilobytes. The GDMS will automatically cleanup temporary files as required. In practise, this means that temporary files from previous sessions are deleted when theGDMS is run. Note that only the GUI implementation cleans the /tmp directory. The GDMS web does not perform this operation as it does not create it's temporary files in this manner. For further information refer to the GDMS web documentation. One limitation which this approach imposes is that is that it restricts the number of concurrent sessions to one on any single machine. The problem is that starting a new session will result in an attempt to clean any files generated by other sessions. The caches are only used during processing and are discarded, however the race condition exists and could cause problems.