Q:I need to implement a backup infrastructure for a handful of remote sites (each with a single server), and I have no IT staff to manage the process at these locations. Instead of a conventional tape backup, would it make sense to replicate the data back to my main site and then back it up with a central tape system?- Nelson Di Marco, Network Administrator Westline Medical Management Alameda, CA
What you're asking is if it is possible to replicate data from a satellite office back to the main site with the intent of creating a centralized copy of all your data.
The quick answer is, "Yes." You could back up the central copy to tape, and in doing so, provide backup services for your remote sites. Before you do so, however, you should consider the following pros and cons of this type of approach.
Zero user intervention-Once you've configured the replication software, the process is automated, requiring no user intervention. A tape backup system, in contrast, typically requires someone to change tapes at given intervals. This opens the door to a potential boatload of problems, including human error and device failure.
Minimum time-lapse to recovery point-Traditional backup software copies data in batches, which means the recovery point is the time of the last backup. Replication software, on the other hand, is often real-time, which means that data is replicated as changes are committed to disk. The recovery point is determined by the latency of the WAN connection. The bottom line: The risk of losing data is less with replication software.
Live copy of the data-Most replication packages allow you to pause the replication process to access replicated data, which means you have immediate access to your data (without having to restore from tape first). This could be very handy, particularly if the remote site went offline.
Data distribution-Many replication products allow you to reverse the data flow and push data out to your remote sites, which means you could use the replication system to distribute files or software updates to off-site locations.
Cost-The cost of replication is pretty cheap relative to tape backup if you have only a few servers to worry about. Replication software starts at about $2,500 per server and goes up from there. WAN links can cost as little as a few hundred dollars per month per site. Storage at the central site can be cheap because it doesn't have to be particularly high perfor ming. A terabyte of RAID storage can run you less than $10,000.
No user interface for individual file restore-One of the main functions of backup software is to facilitate the restore of individual files that may be inadvertently deleted or overwritten. Backup software presents a version history for each file and point-and-click tools for managing the restore of that data. Replication software does not.
WAN latency on restore-If you ever have to restore data, you are faced with the latency of the WAN link. The time needed for a complete restore might be prohibitive.
No application specific tools-Many backup software products have special modules for interfacing with e-mail and database systems. For instance, the backup software might be able to administer the restore of individual mailboxes or database tables. Replication software doesn't have such tools.
The bad gets replicated with the good-Replication software replicates all data, including corrupted data and file deletions. So, if your database is somehow compromised, so is your replica. Similarly, if you delete a vital file, it will be deleted from the replica. (Note: Some replication products can be configured so that file deletions are not duplicated on the target.)
Multi-platform support-None of the leading host-based replication products support multiple simultaneous platforms. So, if you have a mix of platforms, you may need multiple replication systems.
THE BEST OF BOTH WORLDS
There are some solutions on the market that combine the best of replication and backup technologies. These products replicate files to a central server in real-time. The server then tracks file versions and facilitates the restore of individual files.
These solutions are usually limited by the platforms they support and/or the amount of data they can manage. They are typically offered by SSPs in the form of remote backup services, not as off-the-shelf products. I am aware of at least one vendor who is developing multi-platform enterprise solutions based on this concept.
The decision to replicate data should be carefully weighed. If your applications are simple, if downtime isn't a huge problem, and you don't have the people resources, replication is the way to go. If you need greater granularity or need to restore files quickly, you will need a conventional backup system.
However, if your conventional backups are failing, then replication software is a wonderful alternative to nothing!
If you have a question you'd like to ask our experts, please e-mail Heidi Biggar at firstname.lastname@example.org.
Jacob Farmer is the CTO of Cambridge Computer Services. He can be reached at jacobf@cambridge computer.com.