Email server at end of life

A series of server problems with the Cyrus mail server have deprived students, faculty, and staff of access to email for several hours at a time over the past month. Access to email has been interrupted on several occasions, including last Wednesday when Computing Services announced mail service would be unavailable from 11:30 pm to 2:30 am.

According to Karen Van Dusen, departmental computing coordinator in Computing Services, the mail server was taken down last month for brief periods in order to make emergency upgrades to the Cyrus software system. The outage last week was due to these same upgrades to the mail server’s hardware.

The single Cyrus server, which is responsible for handling all incoming mail, has become overwhelmed by the amount of email traffic produced each day. The immense stress on the mail server has caused connections to time out, making it difficult to check email. Considering that there is only one computer to handle the voracious email needs of Carnegie Mellon students and staff, it is understandable that the server has begun to choke under the stress. The problem seems only to be getting worse as the amount of email traffic on campus has increased from three years ago — leading to much greater demands on the server.

The first two email outages last month occurred when the server was disconnected in order to upgrade the Cyrus software. These upgrades implemented optimizations of the Cyrus software in an attempt to improve the server’s performance. Unfortunately, the software improvements were insufficient and necessitated last week’s hardware upgrade.

All of the outages occurred around midnight, affecting students burning the late night oil. However, according to Van Dusen, this timing was necessary because it would be risky to perform a major upgrade at 5 or 6 in the morning — if anything went wrong with the improvement, email would be inaccessible during peak morning hours. “We’re not happy about upgrading in the middle of the week,” said Van Dusen regarding last week’s emergency outage on Wednesday.

While the recent hardware upgrade is expected to improve matters, the server is still in trouble. The ever-increasing email usage on campus will soon tax the server’s capabilities once again. Computing Services is currently working on an “aggregator,” which would allow the mail system to be distributed across several computers.

Additional servers would be installed and access to email would be distributed across the servers.

For example, access might be divided so that each server would serve email for a 20 percent segment of the users.

Computing Services is planning to begin beta testing of the new system in early spring and hopes to have the system in place by next school year.