April 2006

The purpose of this article is to spell out the what and how of Eckerd's co-location project. At the end of this article you will find the definitions of some terms you may not be familiar with.

The core of Eckerd's co-location is SAN-based replication to "dark" equipment in the Alternate Datacenter. We have two identical Fiber-Channel RAIDS, one in each location, with a 50mb network connection between the sites (an "OC-3"). As changes are made to files on the core servers the RAID in the campus datacenter stores the change. As soon as it can, it tells the RAID at the alternate Datacenter to update those records.

When the main campus goes down due to an untoward event, or when ITS decides it is necessary, the RAID in the alternate Datacenter becomes the primary, the fail-over servers mount the volumes that had just been connected to the servers here, and go on serving out web pages and E-mail. Here are some services you use that are handled by SAN-based replication:

• E-Mail, including WebMail, POP, IMAP, and SMTP
• Web pages on home.eckerd.edu, cgi.eckerd.edu and webct.eckerd.edu
• The databases behind Banner, WebMail, and many, many web pages
• Library remote-database access

The web pages on www.eckerd.edu, intranet.eckerd.edu, and academics.eckerd.edu are also replicated by an application called rsync. Because the web pages are largely "read only" when accessed over the web we can have a "Hot" solution here: If main campus is running but something has happened to our web server "farm," the alternate datacenter will pick up the slack.
Some web servers that are hot and ready:

• www.eckerd.edu
• intranet.eckerd.edu
• my.eckerd.edu
• cgi.eckerd.edu
  
• academics.eckerd.edu
  
• TouchNet Payment Gateway (Electronic Payments and Billing)

Definitions:

Fail-Over - passing the work of answering your requests from one server or network to another after a failure.

Hot and Cold Sites - A Hot site is up and running while the main site is, and can even share some of the normal load of the main campus servers. A Cold site is off until the main site goes down.

Storage Area Network (SAN) based Replication - Data is copied in real-time or near real-time - between the devices that store the data on campus and in the alternate data-center. Nothing special has to run on the server that is connected to the SAN.

Operating System (OS)-Based Replication - Data is copied by programs on the server, either in near real-time or on a schedule ("batch copy").

Article by: Walter Moore