Yesterday, University registrar David S. Yeh took full responsibility for difficulties students faced during CoursEnroll on Monday.
“I really take responsibility,” Yeh said. “No one is at fault.”
“I’m disappointed, and I think we could have done a better job,” he said.
It now appears that the delays and error messages that began at 7:30 a.m. and lasted until 11 a.m. were the result of an innocent mistake made in Cornell Information Technologies’ systems operations situation room, according to Rick MacDonald ’71, director of systems and operations at CIT.
At 7:30 a.m. on Monday, there was an instant spike in usage that reached 100-percent capacity in the mainframe, which stores the personal information found in Just the Facts, according to MacDonald. As soon as CoursEnroll had begun, students were waiting for a response from the system.
“During CoursEnroll, we expect that at the instant that the green lights go on that the system is going to be hard-pressed because so many students want to get in as soon as they can and register for their courses,” he said.
In past years, the system had not reached 100 percent so quickly, and this concerned systems operators at CIT. This concern led them to think that service had begun to be delayed due to “looping,” a phenomenon which causes a computer to perform the same set of functions repeatedly.
At 7:47 a.m., systems programmers took the necessary action to correct that problem by rebooting the service, known as Mandarin.
“They had only seen [100-percent utilization of the mainframe] when they had seen looping conditions,” MacDonald said.
This was “somebody making the best analysis they could with the data made available,” but “instead of making things better, it made things worse.”
In a postmortem analysis of what happened, MacDonald said that “in fact, there was nothing wrong with the services.”
Temporarily shutting down Mandarin took CoursEnroll’s connection to the mainframe completely down for about five minutes.
When students tried to submit requests during this period, an error message appeared on their screens.
“The person who took the action did not do anything wrong,” Yeh said. “I’m sure that the person who acted feels badly,” he added. “But from where I’m sitting, I can’t second-guess [that person’s judgment].”
Yeh said second-guessing that judgment might make a future situation worse.
“Next time, they’ll be afraid someone will jump on their neck,” he said. “It’s not about pointing fingers,” he added later.
When the systems came back online, long queues of student requests slowed the system to a crawl. By 11 a.m., those queues disappeared and the system returned to normal operations.
Dave Turner ’05 tried to access CoursEnroll and said he thought things were working faster at first than they had in previous years.
“The initial part went quicker,” he said. “Right at the beginning it was a better performance.”
Turner said he was surprised that CIT had not recognized the problem when seniors enrolled.
“Why don’t they have one class go at a time?” he asked.
Unlike seniors and freshmen, juniors and sophomores have the same pre-enrollment period.
MacDonald said that “there have been volume-related problems in the past. The problems basically always occur in the first instant [of CoursEnroll] … in the rush to get that class. We haven’t seen problems like that in probably a year and a half.”
Turner says he did get into all of his admittedly “strange” courses.
“I’m not trying to bash CoursEnroll. It’s nice to get real-time results,” he said.
Friends at other universities have to wait two weeks for enrollment results. However, he said CoursEnroll “definitely has its flaws.”
Yesterday, The Sun incorrectly reported that the CIT systems engineers acted without knowledge that CoursEnroll was in operation. That information was obtained from Yeh. According to MacDonald, misinformation had been circulating at the time.
Archived article by Peter Norlander