Back It Up

This story keeps me up at night. The Baylor University College of Medicine lost 25 years of research data to Hurricane Allison.

The AID is valuable for many, many reasons. One of which is that the data contained cannot be reproduced. Often one can (painfully) rerun an experiment to reproduce lost data. But when that experiment is time-based and focused on a laboratory light years away, one doesn't have that luxury. Without a Tardis or a '81 DeLorean, we can't go back and time to see what a star was doing in the 20th century. So how do we protect your data?

First, the AID is hosted on a RAID-5 system. This protects it from a single hard-drive failure (which has occured twice since we implemented the first RAID system). If a hard-drive fails, the system continues to operate but we are warned and can swap out the drive without having to take down the system.

Next, every day the AID is backed up to another location on a non-RAID disk. This backup is rotated every six days, meaning we have full backups for each of the previous six days. This is protection against more catastrophic hardware failure or software/user mistakes.

Twice a week, the database is backed up to an external hard-drive. One is taken home by a staff member. This protects the data from theft, fire or other damage to AAVSO HQ. These drives are rotated every four weeks, meaning we have a backup for each of the last four weeks, with the most recent backup residing off-site.

At each meeting of the AAVSO, a copy of the database is burned onto a DVD and given to the current AAVSO president to take back to their home location. This provides a backup every six months that resides far away from AAVSO HQ and Cambridge, MA. The backup is in MySQL format so is easily accessible to those who may need to restore it.

In addition to backups, a transaction log of the database is kept permanently. This log records all changes made to the database from any source. The log is mainly used to troubleshoot issues with the database, but can also be used to reconstruct the database from scratch.

Very similar backups are maintained of the membership database, except they are encrypted and treated more sensitively due to privacy concerns. Thus, I won't go into the exact details about how that database is backed up.

Our hope is that no matter what mother nature of reality throws at us, your 100+ years of variable star data will always be accessible to the community.