If a person is administering a relatively large backup system, the activities in the following list are performed all the time. How easy is it to do these things with the product you are considering?
Making duplicate volumes for off-site storage
-
This feature allows an administrator to make copies of volumes and send the copies off-site. (Actually, a better method is to send the originals off-site. Doing the day-to-day restores from the copies is a constant verification of the copy process. Unfortunately, many products won't allow copies to be made this way.) The right way to do this is to copy from volume to volume, instead of running another backup. The copied volume should have its own identity. That way, the copied volume can be tracked in the catalog. (Older methods of copying volumes resulted in two copies of the same exact volume. The second volume never actually existed in the product's database; it was just another copy of the first volume. The same is true of some VTLs that replicate their data.) Some products also allow an administrator to specify the location of the copy volumes so that he knows which volumes are on-site and which volumes are off-site. That way, the software won't ask for the volume that it knows is off-site, if it knows it has another volume on-site that has the same data on it.
Copying an individual backup
-
There is a downside to copying volumes. Depending on the level of volatility of the data and the luck of the draw, the system may or may not fill up a complete volume each night. If copies are being made and sent off-site every day, what happens to the volume that is half full? If it is copied and sent off-site, the software cannot continue to write to the original volume, even though it has more room available. If it did, it would have to copy the entire volume again. Wouldn't it be better if the software knew how to copy last night's backups, regardless of which volumes happened to receive them?
Individual backup copying does just that. It is the ability to say, "take all of the files that got backed up last night (the backups) and put them on these volumes over here, which I will send off-site." Depending on the complexity of the software, last night's backups may have ended up as 100 MB pieces on 20 volumes. Backup cloning would take those 20 separate 100 MB backups and put them all on one 2,000 MB (2 GB) volume. This would be done after the backup, and the data traffic would stay local to the backup systemso it wouldn't affect the other CPUs, the network, or the users.
Simultaneous copying of backups
-
A very new feature that has begun to appear is the ability to copy the backup as it is being made. If things are set up right, and the software does what it is supposed to do, you actually can make copies in no more time than it takes to make the first backup. The better products would allow the two backup volumes to be different sizes, allowing for differences in compression and different types of media. (Perhaps you may want to put your on-site backups on one type of media and your off-site backups on another type of media.)
Consolidating a host's backup
-
This nifty feature of some backup programs allows for the consolidation of all of the backups for a given host onto a volume or small set of volumes. This makes preparing for a disaster much easier.
Consolidating volumes that have very little data on them
-
This is a volume utilization feature. Due to different expiration times of different backups, there may be a large set of volumes, each of which has only a few hundred megabytes of data that are still useful. Volume consolidation consolidates these backups from all of the partially used volumes onto one volume, allowing the source volumes to be reused.
Filesystem/database discovery
-
Will this product automatically discover all filesystems/drives, or do you have to specify them manually? The latter requires you to constantly update the list of filesystems to be backed upnot good. Also see if it supports this feature on any databases. It sure is nice not to have to maintain the list of SQL Server databases to be backed up.
Excluding files
-
There are always files that should be excluded from backups: temporary Internet files, files in /tmp, spool files, core files, etc. The types of files that need to be excluded vary from site to site. Be aware that different products have different methods for excluding files.
Interface types
-
Most products have Windows, Java, HTML, Motif, Curses, and/or command-line interfaces. Only a few have native Mac interfaces. Make sure the product has an interface that is appropriate for your environment. (I also should state the importance of an optional command-line interface in almost any environment; you can't always count on a graphical display during a disaster recovery, and you also may wish to do your work over a remote connection.)
Notification
-
How should the administration team be notified when a backup fails or needs attention? There are a number of different methods. The most common method is email. Some vendors even allow different email recipients for different types of messages. Other vendors allow you to write custom interface programs that send reports wherever you would like. Some backup products can even send alpha pages if there is a modem available.
Perhaps the most sophisticated notification method is the use of an SNMP trap. SNMP stands for Simple Network Management Protocol and was originally developed for network hardware so that routers from different vendors could understand one another. SNMP has been expanded to include all types of network monitoring, and it is what most of the commercial network monitoring tools are based on.
Installing
-
Most server installations are fairly automated. The real bear is the installation of client-side software. There are usually two ways that client-side software can be installed: manually or automatically. When using the manual method, you first must get the client software to the machine. This is done either by ftp, nfs, or CIFS. Then, run the installation program. Some products can completely automate client installation under certain circumstances.
Upgrading
-
The product is installed once, but it will be upgraded once or twice a year. Therefore, ease of upgrading is another very important feature to consider. Upgrading also may be performed manually or automatically. However, some products (once installed) are able to update the software automatically. Since updating the software on hundreds of clients is the most difficult repetitive task you will do, this feature is quite handy.
