AWS CloudFront for Emergency Web Service

As part of the University's disaster preparedness efforts, key portions of the content of the official University web pages at www.columbia.edu are copied daily to a group of server hosted by AWS CloudFront. This failover service supplements our replicated web servers and high reliability file servers.

The primary purpose of the web backup service is to maintain a web presence in the event of a major disaster and to allow the University to communicate via the web with the rest of the world, including our students, faculty, staff, and their colleagues, friends and families. However, even a 5-minute network or web server failure will cause the AWS CloudFront server to be used.

What's Mirrored at AWS CloudFront?

The Columbia home page and the pages on www.columbia.edu that are linked directly from the home page, plus some of our more popular web pages are copied to AWS CloudFront servers in other parts of the country each morning at 5:00 AM. In the event of a major incident that requires updates to off-campus copies of the the web pages CUIT staff will work with other University officials to update the content on the AWS CloudFront servers. We are also working on a related project to develop an emergency web bulletin board system to enable limited two-way communication when email systems are unavailable.

Which Pages are Not Mirrored at AWS CloudFront?

A significant amount of the University web content is not copied to AWS CloudFront; the service is expensive and our use of it is meant only to protect highly time-critical information that requires widespread, reliable dissemination. Furthermore, only static content can be mirrored.

Specifically, the following types of content and web sites are not mirrored:

  • Web sites that are not hosted on the central CUIT web servers. These include a number of school and departmental sites that have been outsourced or are run by the school, department, or other unit. (If you would like your website hosted by CUIT, please submit a ticket to make a request.)
  • Dynamically-generated content such as search pages are generated dynamically using a search engine behind the scenes.
  • Web-based applications. These pages are dynamic and require back-end application servers, databases which cannot be mirrored at AWS CloudFront.
  • Web pages and applications not run on CUIT servers such as Student Services Online.
  • Any secure web pages on the CUIT-run secure web servers (HTTPS).

How Do I Know If My Files are Mirrored?

The mirror server is named http://www.columbia.edu.s3-website-us-east-1.amazonaws.com/. Point your browser there to test it. Please note that during normal operation, any pages you author that link explicitly to http://www.columbia.edu instead of using a relative URL will go back to the CUIT web server and will not be a proper test. When the main server is considered down by AWS CloudFront, these links will also be redirected to the fail-over server at AWS CloudFront. To test these links, either make them relative URLs in your web pages or explicitly test them by hand on the fail-over server. To make a relative URL use, for example,

<a href="/cu/dept/sample.html">

instead of

<a href="http://www.columbia.edu/cu/dept/sample.html">

Why am I Connected to AWS CloudFront When I can Still Reach Columbia?

The AWS CloudFront fail-over service works automatically and can be triggered by even relatively minor short-duration Internet connectivity outages. See "How it works", below.

You can tell your browser is on the backup server because AWS CloudFront's server sends you a redirect to http://www.columbia.edu.s3-website-us-east-1.amazonaws.com/. This appears in the URL text box. This is actually not the optimal way for this to work since references to relative URL's will also stick on the fail-over server.

To get "un-stuck" you need to re-reference www.columbia.edu and may need to restart your browser. This is because the IP address of www.columbia.edu that points at the fail-over server is cached by some browsers. This is a bug in the browser and/or operating system but it's not like we have much influence over Microsoft to fix this for deployed browsers!

How Does My Organization Add Pages to the Mirror?

If you have critical pages that are not currently being copied to the fail-over mirror, please submit a ticket.