When exporting large spaces as PDF, there's a big performance issue. When Confluence starts converting the collected pages to PDF, the application becomes unavailable for the users. After some time, there's a java.lang.OutOfMemoryError: Java heap space in the atlassian-confluence.log.

The OutOfMemoryError problem can be fixed by increasing the Xmx from 1024m to 2048m.

There some options though, like:

1. Possibility of running the PDF export at off-hours, on a synchronized job.
2. Attachment Size. Limiting it by size and type, finding large attachments.
3. Pages not included in the PDF be moved to a separate space (also relevant for (1) above).