Evolving backup: From tape to disk to cloud or none at all?

The backup market continues to expand with new options and capabilities, despite the fact that organizations tend to be pretty conservative when it comes to making large scale changes to backup infrastructure and processes. These backup options come with new processes and opportunities that can be exploited for improved disaster recovery or streamlined operations.

In a September 2010 TechRepublic article, I discussed Westminster College’s pending migration from Backup Exec to Microsoft’s Data Protection Manager 2010 and, more recently, I provided an update to that article that outlines where we are now.

However, with the move from a 100 percent tape-based backup infrastructure to a disk-to-disk-to-tape system, we’ve only made the next logical step in the backup and recovery process; there are still lots of options at our disposal, but not all of them fit our needs.

In this article, I discuss other data protection options I’ve seen and heard about and provide my thoughts about each one.

Tape-only backups
Yesterday’s legacy tape-only backup mechanisms are becoming obsolete, as organizations look for different ways to maintain highly available infrastructures that are also able to withstand disasters. I admit that Westminster has been behind the curve a bit in our backup/recovery processes, but we’d been served well by those processes until only recently.

We recognized the significant drawbacks that tape-only backups provide for growing organizations, including:

• Limited opportunity for backup. Tape is slow. It’s common to see once per day incremental or differential backups that follow a once per week full backups. Organizations don’t want to see backup operations significantly impact production resources, so these backup operations are pushed to the wee hours of the morning and to weekends.
• Slow recovery. Recovering from a tape backup is no fun. It’s slow, error prone, and can be difficult, depending on the number of tapes from which data has to be recovered.
• Equipment challenges. It’s not uncommon for tapes to be usable only in the drives that wrote the data in the first place with the software that did it. So, recovering with a newer tape unit might not be possible.

Disk-only backups
The move to a disk-based backup and recovery architecture makes a lot of sense when you consider the potentially huge benefits, including easy expansion of available disk space, speed increases that result in more continuous protection, and very fast recovery.

The primary downside to a disk-only backup architecture is the possibility of failure. Organizations are not as likely to place the same availability importance on the backup system, so it’s more prone to failure than other production systems.

If a backup system suffers a catastrophic failure that results in the loss of all protected data, an organization is left at significant risk. This is why I strongly believe that disk-based backup systems need to be augmented with tape-based or cloud-based backup systems that add an additional layer of protection to the mix.

Although the addition of disk to a tape-only backup infrastructure isn’t a silver bullet, it can solve, in many instances, the slow backup and recovery problems that plague tape-only installations. Disk is much faster than tape. When coupled with tape, organizations can get the best of both worlds–quick backup and quick short-term recovery, along with long-term backup to tape that doesn’t impact production systems.

The same physical equipment challenges around tape remain, but these challenges are often worth it to add another layer of protection to the environment.

Cloud-based backups
In a July 2010 TechRepublic article, I outlined five challenges that organizations will face when moving backup to the cloud, which are: cost, insufficient bandwidth, security issues, recovery challenges due to the aforementioned bandwidth shortage, and the reliability of the vendors providing the service.

The benefits of cloud-based backups include:

• Simplicity. Many cloud providers offer a "set it ad forget it" operation when it comes to data protection.
• Reduced infrastructure complexity. In theory, moving to an outsourced provider can reduce how much equipment is necessary to achieve a goal. This would include a backup service, depending on how that service is used.
• An additional option or tier. Cloud backup is not likely to fully supplant on-premises solutions, particularly for medium and large businesses. It does, however, present an additional option or protection layer that did not used to be available.

Application-provided protection
I’m not going to describe all of the possible application-level protection possibilities out there; instead, I’ll focus on one option that I find particularly interesting: Exchange 2010 Database Availability Groups (DAGs). On the surface, DAGs are simply an availability mechanism, but then again, that’s sort of the role that backups play, too.

Microsoft IT has taken DAGs to a whole new level and made the decision to implement DAGs and eschew traditional Exchange backups altogether. Here’s an excerpt from the document entitled Exchange Server 2010 Design and Architecture at Microsoft:

Decrease costs. This included redesigning server architectures and backup solutions for high availability to meet challenging SLAs. In redesigning server architectures, Microsoft IT heavily focused on incorporating the features directly available in Exchange Server 2010, replacing continuous cluster replication (CCR)-based Mailbox server clusters with the new database availability groups (DAGs), and eliminating backups altogether. All of these considerations resulted in significant cost savings.

In short, Microsoft got rid of its backup infrastructure by carefully implementing DAGs to replace the role that was formerly occupied by backup software, hardware, and processes.

I’ll be surprised if I see a lot of companies eliminating traditional backups in favor of other availability mechanisms. I tend to look at availability mechanisms such as DAG as more dynamic protection schemes, while traditional backups offer some kind of "static expectation" on which I can depend. Dynamic and truly continuous is good, but I find that I like the safety that comes with some kind of point-in-time backup, even if it’s being taken as often as every 15 minutes.

Summary
The line between backup and availability is somewhat blurry, as evidenced by Microsoft IT’s decision to eliminate backups. The choice about what to use for backup depends on the organization’s needs and may include just one traditional option, such as tape, or it might continue to rely on local infrastructure that is then augmented with a cloud layer that provides additional protection.