Backup and verification operations can create significant increases in network traffic. This can lead to reduced performance of your servers and all of the segments of the network involved in the backup. One solution to the problem of course is scheduling backups at a time when network usage is low, like before or after hours. Since this may not always be possible, you can perform backups in smaller incremental steps during normal hours.
It is possible to create an isolated network segment just for backups, a somewhat costly but effective alternative. This works well in an environment where most of the data to be backed up resides on centralized servers. Differential backups save some time, since they backup files that have changed since the last full backup. They are essentially the same as incremental backups but they don't turn off the archive bit on the files they back up. Successive differential backups become longer and longer as more files have changed. A good advantage to this type of backup is that if a restore is needed, it requires accessing only the latest full backup and the latest differential backup; none of the interim backups are needed. Some backup programs come with both incremental and differential backup types. A daily backup or daily copy, as it is often called, is the only backup type that takes only the file modification date into account when it is deciding what to back up. It backs up only that day's changes. It can be a time saver on the network if it is performed religiously every day. Otherwise, you may miss some changes made after the last backup and before the current one.
If the servers are fairly close to each other physically, you can add a second network adapter to each server, along with an additional hub and the wiring needed to complete an isolated network segment. When you run backup with this new segment, there will be no significant impact on network performance. The additional disk and network input/output may affect the performance of individual servers somewhat, but the network won't be clogged by backup data.
When performing backups over a network, you should create network connections from the computer with the backup drive to the computers you want to back up. The computer with the backup drive acts as a client, while the computers being backed up act as servers. Be sure that the remote computers are configured that way. While there is no optimal way to cut traffic on the network during normal hours, the only other option is to resort to after-hours backups.
There are many other technique that can help you substantially reduce network backup traffic. Consider using snapshot technology to create point-in-time copies of your data. With snapshots, changes made to your file system after you have backed up your system onto tape or disk for the full backup are stored as metadata in subsequent incremental backups.
Another useful technology for eliminating network backup traffic in large data sets is to invest in storage area networks and serverless or LAN-free backup technology. Backup software stores backup information as metadata, and instructs the storage servers how to perform the backup directly in-band (over Fibre Channel or the storage network) directly to the backup media. Many of the newer enterprise backup programs like VERITAS Netbackup and Legato Networker support serverless backups.
Whatever method you use, regular tested backups are your last and best defense in protecting your data.
Barrie Sosinsky ([email protected])is president of consulting company Sosinsky and Associates (Medfield MA). He has written extensively on a variety of computer topics. His company specializes in custom software (database and Web related), training and technical documentation.
Unix Backup and Recovery
Author: W. Curtis Preston
Publisher: O'Reilly & Associates
Unix Backup and Recovery provides a complete overview of all facets of Unix backup and recovery, and offers practical, affordable backup and recovery solutions for environments of all sizes and budgets. The book begins with detailed explanations of the native backup utilities available to the Unix administrator, and ends with practical advice on choosing a commercial backup utility.