For some companies, having a database down means being out of business. In these cases, time to restore is almost
as critical as being able to restore at all.
In the case of YP.Net, a leading online yellow pages company, the database literally is the business. Users visit the site millions of times a day to look up businesses in the United States and Canada. In the event of a problem, the company has to get the database back as quickly as possible.
In the next year, the company will add co-location sites in Nevada and Florida to its Mesa, AZ, facility with failover between all the sites for disaster recovery. Meantime, and for day-to-day problems, the company relies on a combination of delta backups and a three-tier backup architecture.
YP.Net Chief Technology officer John Raven explained that the company's backup choices were strongly influenced by two other consideration. One is cost -- the need achieve acceptable levels of protection as cheaply as possible. The other is that the YP.Net database is essentially read-only. While the other kinds of data, such as programmers' work and general office files, change constantly, the database is only updated on a regular, fairly infrequent, schedule.
The relatively small amount of changes to the 4 terabytes of corporate data allow allows the company to rely heavily on delta snapshot backups where only the data which has been changed is backed up. The delta backups go to a disk-based near-line storage system, so data can be restored in a matter of seconds if needed. Alternatively, the entire site can be run off nearline storage, albeit at a cost in performance. Data is further backed up to tape for offline storage and copies of the tapes are stored offsite.
Because cost is also important, there are significant differences between the online and nearline storage systems. The online system is highly optimized for fast response to queries and is built around relatively small (18 gig), high-speed, disks to maximize performance. The near-line system is built using much cheaper disks connected by IDE and Serial ATA.
For more information:Tip: Using NAS for database storage
Tip: The problem backing up big databases
Tip: Snapshots in SQL Server 2000
About the author: Rick Cook has been writing about mass storage since the days when the term meant an 80K floppy disk. The computers he learned on used ferrite cores and magnetic drums. For the last twenty years he has been a freelance writer specializing in storage and other computer issues.