Feds brace Web sites for Starr report traffic

Federal computer specialists last week struggled to prepare World Wide Web sites to manage an onslaught from the public trying to access the hottest document in Internet history— Independent Counsel Kenneth Starr's report to Congress on President Clinton.

Hours before the expected release of the 445-page document Sept. 11, federal Web sites that planned to carry the report were already clogged with traffic.

The House of Representatives and congressional agencies, including the Library of Congress and the Government Printing Office, offered the Starr report online at four congressional sites— thomas.loc.gov/icreport, icreport.house.gov/icreport, icreport.house.gov/press and www.access.gpo.gov/congress/icreport.

But at about 5 p.m. Sept. 11— more than an hour after the report was released— the Thomas and GPO sites showed they were managing the flood of requests. The average round-trip time for packets of data took about 76 milliseconds for the Thomas site and 104 milliseconds for the GPO site, according to the Internet Weather Report, operated by Clear Ink Corp., Walnut Creek, Calif.

"We're looking pretty good," said T.C. Evans, GPO's assistant director for the Office of Electronic Dissemination Services and who helped coordinate online release of the Starr report.

Evans said GPO publicized the hosting of the report on new servers the office had dedicated exclusively to the Starr report, at icreport.access.gpo.gov/report/1cover.htm.

At FCW's offices, reporters trying to access the GPO site experienced more frequent and longer delays as the afternoon wore on. Messages such as "temporarily unavailable" became more common.

Some response problems also may have been attributable to a major Internet outage caused by a train derailment outside Atlanta that cut a number of key fiber-optic cables.

At the LOC, which operates the Thomas site, Herbert Becker, director of information technology services, said that shortly after 5 p.m. traffic for the report was "very, very heavy" on a main LOC server and on another LOC server dedicated specifically to the Starr report.

"There are a lot of people getting in, but there are a lot of people timing out," he said. He downplayed the notion that heavy traffic could crash the server. Rather, heavy traffic should simply mean there are more people who cannot get immediate access to the report and who have to wait, he said.

Diverting the amount of traffic on federal Web sites were commercial Web sites that carried the Starr report as well. Media outlets, such as CNN and CNBC, as well as technology companies, including Netscape Communications Corp., offered relief by posting the report on their sites and suggesting that federal coordinators of the project include links to them.

A Netscape spokesman said his company had planned to carry the report on its Web site, with a link from home.netscape.com, which is the default starting page for users of Netscape's Navigator Web browser.

The voluminous Starr report was much sought after by the public, creating the potential for online logjams for congressional sites that posted the document. People involved in the project last week expressed uncertainty over whether federal systems could painlessly accommodate all the hits from the curious public.

"There's going to certainly be quite a few people downloading the entire version, and that's going to put a strain on any system," GPO's Evans said before the report was posted. Evans helped to coordinate the online release of the Starr report.

As of early on Sept. 11, Evans and others working on the project had been unsure of the format of the Starr report— a crucial detail for how the document would be presented online. Posting several, smaller online portions of the document instead of one large online document would alleviate delays for visitors trying to access the information.

But all Evans knew the morning of Sept. 11 was that he could get a copy of the Starr report in electronic form and that the report would be broken down into at least three sections: a 25-page introduction, a 280-page narrative and a 140-page section covering grounds for impeachment.

In the end, the electronic report provided more than 200 "definable breaks" that Web surfers could click on to read a portion of the Starr report.

"If they break it up into small pieces, almost any server could handle it," said Carlynn Thompson, director for research, development and acquisition information support at the Defense Technical Information Center. "The things you have to worry about are bandwidth and breaking it up into chunks." She said if a document is broken up well, the servers supporting it can handle millions of hits per day.

As for acquiring bandwidth for the report, congressional agencies braced quickly and feverishly. The LOC's Thomas Web site connects to the Internet through a T-3 (45 megabits/sec) line supplied by Data Research Associates, a small St. Louis-based Internet service provider that focuses on marketing networking services to libraries. Last week, the LOC asked Data Research for additional circuit capacity to handle the anticipated traffic spike for the Starr report.

Joe Bonwhich, a Data Research vice president, declined to comment on the amount of extra bandwidth the company supplied, but he said that starting Sept. 9 "our personnel have been working feverishly to handle the load."

Another factor that could have slowed down access to the document was the format of the document.

Evans said GPO would make the report available on two dedicated servers as plain ASCII text in Hypertext Markup Language as well as in portable document format (PDF), which displays images of pages as they appear in their original form so that Web visitors can see the document as members of Congress will see it. But PDFs usually take more time to load than text files because PDF files include more data.

The Internet release of Starr's report on his investigation into possible presidential wrong-doing is not the first time presidential news has caused a rush on the Web.

Bryan Pfaffenberger, a professor of technology, culture and communications at the University of Virginia, said Web broadcasting of Clinton's inauguration in 1996 endured close to 3,000 hits per second and bogged down the Internet backbone on the East Coast.

But for Pfaffenberger, the release of the Starr report on the Web represented more than a simple bogging down of a backbone.

"I'm terrified of an Internet meltdown today," he said on the day the report was released. "The number of people who are likely to attempt to access this thing is potentially huge. I'm planning to download it."