Main Nav


If our main web site goes down at Princeton we have a failover web site with a simple "Technical difficulties" page on a virtual server at another university that will answer requests. We use a service from UltraDNS to handle the DNS failover for the IP addresses to shift from our Princeton network to the emergency site.


We are looking to re-architect this setup and are curious as to what other configurations people have in place in case their primary web site has a failure. For example, we are looking at the F5 global load balancing options and other hosted solutions. Any insights would be much appreciated!


Thank you,


Joe Karam

Manager, Collaboration Services Group


********** Participation and subscription information for this EDUCAUSE Constituent Group discussion list can be found at


Message from


At Georgetown our current setup is, briefly:
  • For off-campus requests, the external DNS provider (UltraDNS) directs requests to a load balancer in one of two off-campus commercial data centers, in different parts of the country
  • Behind each off-campus load balancer are a couple of Linux/Apache servers serving static web content
  • For on-campus requests, there's another web server at the main university data center and at the backup data center, so that on-campus users are served if the university's Internet connections are down, and so that on-campus users can potentially be served different content from the general public in the event of an on-campus emergency

The enterprise CMS pushes static web content out to the various web servers; there is a backup plan for SFTPing out content in case the CMS is down. Content includes the university top tier web site and various other sites that are deemed "high availability" -- that is, critical for emergency communications and basic business continuity. The university home page can be switched over to a pre-designed alternate version that prioritizes emergency communications content over regular day-to-day navigation.

Of course, only a fraction of Georgetown's total web content is included in this setup. To maximize reliability it doesn't include any server-side applications (obviously there can be client-side coding within the static content to provide some interactivity).

I haven't been involved in maintaining this since it was implemented around 2008, but I can refer you if you like.



Piet Niederhausen
Web and Data Architect
Georgetown University Information Services