The design of the www.washignton.edu cluster prevents data files from being manipulated to prevent content corruption. However, this also makes it not possible to save persistent data in the filesystem. In addition, since there are multiple web servers, any data saved on one server cannot be seen by the other servers.
The gather directories are a method by which you can save data on the web servers which can later be collected by a batch process.
This case study shows a CGI which will collect the browser and operating system information of each connection and save it in the server's gather directory. Each evening a batch process runs to collect this data and summarize into a data file.