++ to batch processing.

The way I would do it is to have a simple CGI script on the server that sets a flag to indicate that a dump has been requested by writing an empty file to a certain place. (After checking that the requesting user is authenitcated, and has the right to request a dump.)

Then have a cron job that runs every few minutes as a more privalaged user. If the flag is present it can do the dump and write it to a location that the web server can read. If there is no flag, or the last dump was to recent then it exits immedately.

From the end user's point of view the process is to visit a special web page, enter a username and password, then wait 5 minutes or so before downloading the result.