Spam and Virus Filtering at Scale

I am currently working on a large-scale email system designed to serve millions of mailboxes — something comparable to Hotmail. Beyond the basic ability to send and receive messages, the most critical feature is spam and virus filtering.

For spam filtering, I am using SpamBayes, which has already proven itself on our smaller production servers. For virus filtering, the setup combines Anomy Sanitizer with ClamAV. Anomy Sanitizer strips or defangs dangerous attachments before they reach the user, while ClamAV scans message bodies and attachments against its virus signature database. Together, the two tools form a solid defense layer, and early results look promising.

I wrote a Chinese document describing how to integrate SpamBayes, Anomy Sanitizer, and ClamAV into an existing email system.

The next challenge is making the whole pipeline both efficient and scalable. Any suggestions?