News server crashes twice due to software failure

The netnews.upenn.edu server -- which provides the University community with access to approximately 7,000 Usenet and Penn newsgroups -- crashed Wednesday night. As of late last night, netnews users could only access newsgroups in the local "upenn.*" hierarchy, which includes about 600 newsgroups. According to Bill Magill, manager of network computing services for Data Communications and Computing Services, netnews first crashed at approximately 8 p.m. Wednesday. Although the server was back on line at about 11:15 p.m., it went down again early Thursday morning. Magill said netnews should be fully restored by early this morning, but added that no "formal deadline" is in place. Noam Arzt, director of information technology and architecture for Information Systems and Computing, attributed the crash to a software failure. Arzt said ISC needs "special software? to manage the very large amount of disk space" required to store every unexpired Usenet article on netnews. "Part of that software failed and needed to be reconfigured," Arzt explained. Additionally, Magill said the continuous increase in news volume has been "astronomical" -- and this helped precipitate the crash. "The growth was even more dramatic than anyone had predicted," he said. Magill said that even though DCCS had finished replacing the entire netnews system in December, its news volume projections were too "conservative." Arzt added that no articles in upenn.* newsgroups were actually erased due to the crash. He said DCCS takes "very different steps to preserve" the upenn.* newsgroups as opposed to those on Usenet. DCCS backs up all upenn.* articles from the current academic year, Arzt explained. However, netnews only has the capacity to retain five days' worth of Usenet articles -- and that alone includes millions of posts from more than 6,000 newsgroups. Arzt said that since netnews deletes Usenet articles that are more than five days old, the crash's effects will be all but invisible by the middle of next week. "Within several days it will have everything it would have had" if the crash had not occurred, Arzt said. DCCS posted updates on the netnews situation to its World Wide Web page throughout the day yesterday. Arzt is confident that this problem will not reoccur in the near future. "It is unlikely that this will happen again because we understand the phenomenon better now," he said. However, Arzt explained that Penn is not the only university to have suffered this type of crash. "It has happened at other institutions using similar servers and similar software," Arzt said, adding that Penn's netnews system has been running on a Digital Equipment Corporation Alpha supercomputer since December 1995. Initially, the crash seemed to confuse users -- several of whom posted to various newsgroups wondering what had happened. But by last night, students did not seem to mind the system failure. College junior Cristobal Cardona said although he subscribes to "about one hundred" newsgroups, the crash did not really affect him. "It's no big deal," he said, adding that "all systems have problems" occasionally. And College sophomore Patrick Danner felt similarly. "It's not going to have a profound impact on my life," he said. Still, Arzt noted that 46 members of the University community were using netnews within a minute of its restart on Wednesday night.

Please note All comments are eligible for publication in The Daily Pennsylvanian.