Upgrade Attempt Slows Network

Students and faculty lost FAS network access across campus yesterday after an early-morning system upgrade went awry.

Network users were unable to check e-mail, surf the Internet or start machines connected to the network intermittently throughout the day. The network was stable by 5:20 p.m., according to FAS Network Operations.

Harvard Arts and Sciences Computer Services (HASCS) administrators said the campus-wide blackout was caused by a glitch in a network upgrade.

HASCS performed several upgrades between 4:30 and 7:30 a.m. yesterday, among them the installation of new software on its core switch, a central part of the network in the Science Center. According to HASCS Director Franklin M. Steen, the new software unexpectedly caused the switch to malfunction.

With the network fading in and out, e-mail was unavailable for much of the day. The shutdown also affected computers that are set up to look for the network as they boot up, including machines that run Microsoft Windows.

"This morning I woke up and tried to boot up my computer," said Thomas J. Mucha '03, who said he encountered a blank screen. "I thought it was my computer's fault."

Rick Osterberg '96, database applications specialist for HASCS, speculated that the intermittent connection to the network may have caused start-up problems that might not have occurred if the network were completely unavailable.

"The instabilities may have been of a nature that some traffic was getting through--but it wasn't a reliable communications stream," Osterberg said. "Sometimes an unreliable stream is worse than no stream at all."

HASCS eventually reinstalled the old software on the core switch, stabilizing the network.

Osterberg said the main focus of the team trying to fix the problems yesterday was to get the network back up, not to determine what went wrong.

"Our priority is to get it working now," he said last night. "We'll figure out what happened tomorrow."

David M. Sobel, manager of user services at HASCS, said such a network interruption was unusual.

"This was very unprecedented. We were all kind of shocked," he said. "For the most part we're very pleased with the reliability of the network. This is the exception, not the rule."

Steen agreed. "People take the network systems for granted because they're so reliable."

The new software that caused the problem was part of a major upgrade program underway at HASCS to make the network faster and safer.

Steen said in the process of upgrading, problems are inevitable.

"To run a network there are always things that work and things that don't work," he said.

Steen said yesterday that while reverting to the old core switch software is not an immediate setback, HASCS will eventually try to install the new version again.

"For the time being, it's not a problem," he said. "There's no question we can't run that forever. Sooner or later we'll have to upgrade."

In the meantime, Steen said HASCS is trying to avoid further disruptions to network availability.

"We apologize for it and ask for patience," he said of yesterday's outage.