Large email systems

This is a list of documents about large email systems, almost all of which have been set up by universities. They have multi-server clusters for load sharing, some for fault tolerance, and all have centralised user authentication databases. Unless I'm mistaken, all of them run on Unix or Linux, and all of them use Cyrus IMAP. Some use Sendmail, many use Posfix, as MTA.

Unless I'm mistaken, none of them are for less than 10,000 users. All of them are single-location clusters, unlike most Merce setups which are multi-locational.

1. "HEC Montreal: Deployment of Large-Scale Mail Installation" by Ludovic Marcotte, http://www.linuxjournal.com/article/7323 [1]
2. "Scaling up Cambridge University's Email Service" by David Carter and Tony Finch,
   http://www-uxsup.csx.cam.ac.uk/~fanf2/hermes/doc/talks/2004-02-ukuug/paper.html
   A large and thorough article. 24,500 user accounts, 7,500 concurrent users typical. They migrated from UW-IMAP to a Cyrus cluster.
3. "CMU's Cyrus IMAP configuration", http://cyrusimap.web.cmu.edu/configuration.html [2]
   This page does not give you IMAP configuration; it gives you hardware configuration and OS specs for the IMAP servers.
4. "Columbia's Cyrus IMAP Move", http://www.columbia.edu/cu/news/05/12/cyrus.html [3]
   A brief news item, more like a press release. But you get a sense of the size of their mail setup: 66,000 users.
5. "Indiana's Cyrus IMAP information", http://uitspress.iu.edu/040505_cyrus.html [4]
   This page could not be found, but I found this [5] instead. This is a hands-on article by the techies who actually did the migration. Lots of information about IMAP migration, LDAP configuration, etc.
6. "Stanford's email system discussion", http://www.stanford.edu/dept/its/vision/email.html [6]
   Quoting from the first paragraph:

   "This document tries to cover the strategic vision for email services at Stanford, focusing specifically on the centralized email services provided by and managed by IT Services. This includes the email gateway services, spam/virus filtering, POP/IMAP/webmail client services, and the mailing list service."

7. "Toward an automated vulnerability comparison of open-source IMAP servers" by Chaos Golubitsky, http://www.usenix.org/events/lisa05/tech/golubitsky/golubitsky.pdf [7]
   This article discusses the relative security level of Courier IMAP, Cyrus IMAP, and UW-IMAP. UW-IMAP is very poor in performance for large installations, hence is not used in real sites other than tiny ones. Courier and Cyrus are both quite popular open source IMAP implementations.

Also see "Cyrus IMAP aggregation", http://cyrusimap.web.cmu.edu/ag.html [8]

It would be really nice if some of you could pull out each document, print-to-file each one, convert to PDF and store the PDF locally on our server, linked to this page. That way, we won't be at the mercy of documents disappearing from the Internet. In case a document is in the form of multiple HTML pages, you can either make local copies of those HTML pages and patch their links to make a consistent sequence of local pages, or you could print-to-PDF each page and then stitch the PDFs together using the pdfjoin utility which comes as part of the pdfjam package on Ubuntu.

All this was obtained from the Linux Journal dated Nov 2007. This issue has an article titled "Building a scalable high-availability e-mail system with Active Directory and more" [9] by Jack Chongjie Xue, and discusses the story of Marshall University.

An interesting aside: Mirapoint [10], one of the world leaders in the hardened email server appliances, has targeted this sector (large colleges and universities) with great success. Education is one of the important verticals they have worked with.