Load Balancing

The 7-year-old , Web site of the Centers for Disease Control and Prevention, has been on what CIO Jim Seligman calls “a continuous growth curve.”The site hosts 120,000 pages and dozens of databases used by the public and health care professionals. Through September 2001 it was averaging 4 million unique visitors each month.“That’s among the top government Web sites,” Seligman said. “Since the anthrax business, usage has jumped.” After counting more than 9.1 million visitors in October, 5.2 million in November and 7 million in December, CDC turned to a managed-content delivery service from Akamai Technologies Inc. of Cambridge, Mass.The site quickly became one of the fastest government sites on the Keynote Government 40 performance index. CDC.gov took the No. 2 spot in December with an average response time of 0.44 seconds. It remained No. 2 for five weeks.“That was a nice by-product,” Seligman said of the ranking from Keynote Systems Inc. of San Mateo, Calif., but he was more interested in sustainability and reliability. By caching data on Akamai servers around the world, Seligman said, CDC could keep its site available in spite of network interruptions, traffic spikes or cyberattacks.The EdgeSuite service redirects visitor requests to Akamai’s Domain Name System server, which sends them to a caching server at the Internet’s edge. CDC’s host servers are polled only for dynamic or updated data. That reduces their workload, improving availability and relieving the need to upgrade and manage a larger infrastructure.The edge servers also help shield CDC data centers from intrusion and absorb traffic from service denial attacks.The Akamai service uses about 13,000 servers on 1,000 networks in 63 countries. It also can provide backup by delivering stored or default pages when a host server is down.“We are an overlay,” Akamai federal sales manager Christopher S. Carlston said. “We sit on the networks that already are there.”CDC links to Akamai’s distributed platform through the agency’s Internet provider. “That’s the only pipe they use to get the information out to our servers,” Carlston said. He described the servers as robust Intel machines running a modified Linux operating system.CDC had been in talks with the company for months before the Sept. 11 attacks and the release of anthrax virus, after which the site became “the de facto reference point for bioterrorism information around the world,” Carlson said.Seligman said CDC gained visibility in the mid-1980s with the spread of the AIDS virus. “Long before 9-11, Akamai had approached us, and we had expressed interest in their technology,” he said. “It was going to happen anyway. Having the content distributed was one particularly appealing part. The other appealing part was sustainability in the face of some kind of problem.”After hardware failure took the CDC Web site down for several hours on Oct. 19, turning on the Akamai service was a matter of initiating the redirect process at the DNS servers.“It was an easy implementation,” Seligman said, “which is quite unusual in the IT realm.”He described it as a partial step to outsourcing the agency’s Web presence. “We still host on our own equipment,” which remains accessible from the Internet, he said. Database searches still occur at CDC’s data centers, but the infrastructure for handling large volumes of traffic is Akamai’s.CDC meanwhile is redesigning the site for easier navigation. The original design reflected agency organization rather than visitors’ needs. The new focus will be on the user experience, Seligman said.