Affordable IT: Storage

When it comes to backing up applications outside the data center, there are multiple strategies. Do you back up to a single, central server? Does each box have its

own backup repository? Do you back up all machines the same, or are there special procedures for critical servers?

We've all heard the disk-to-disk backup "strategy" that goes: "We copied all the documents in the Contracts folder over to a folder named Backup on our server." Even more sophisticated shops that have tape systems in place make the fatal mistake of failing to verify that the backups worked. Trusting the system because you don't have time to do spot checks is risky business.

It doesn't have to be this way. Storage system prices have been dropping. New options and deals are plentiful. For example, Microsoft's introduction in May of Windows Storage Server gave vendors a low-cost base platform, spurring customization. Windows Storage Server consolidates storage functionality in a Windows platform at a reasonable price to OEMs. Meanwhile, S-ATA (Serial ATA) has reduced the per-gigabyte cost of hard drives--S-ATA drives are now considerably less expensive than their SCSI counterparts. All this lets vendors build inexpensive storage machines on a common platform.

The result is that many affordable backup options are available for small and midsize businesses as well as enterprise-level departments. D2D2T (disk-to-disk-to-tape) appliances let backups run to disk and then, over time, stream data off to tape, enabling your core systems to back up using fast disk-to-disk techniques, and off-loading the slower disk-to-tape backups to the appliances. Depending on the technology and vendor you use, disk-to-disk backups can be two to five times faster than disk-to-tape backups. More important, restores in those first few hours between backup and when the data is streamed to tape occur at random-seek disk speeds instead of streaming tape speeds.The benefits of these systems over conventional low-end backup methodologies are compelling. First, the speed advantages mean you lose less network bandwidth and processing time on your servers. Second, most restore requests are resolved quickly. You need only last night's backups to fix this type of problem, and last night's backups should still be sitting on disk--faster restores mean getting the business back up and running quickly. Finally, you can (and should) store tapes off-site, meaning you have a backup even if your entire company suffers a catastrophe (see "Disk, Tape and the Better Backup,").

Most D2D2T appliances work by presenting themselves as tape drives. They accept the commands of major backup software systems just as if they were a tape archive, then stream the associated data to disk. On a regular schedule that you set, they then move the backups off to tape and delete the disk-based copy on a separate schedule, though some delete at the end of the tape backup process. Of course, you still must determine what gets backed up and when, and configure your servers to perform the backups. In addition, you must take the time to move the tapes off-site, as well as check your backups regularly to ensure they're valid.

More and more D2D2T devices are offering iSCSI targets. Whether you prefer iSCSI or CIFS and NFS is up to you, but bear in mind that D2D2T appliances require backups to be over the network. If your backups have been local to your machines, make certain your network can handle the additional traffic before you spend money. For more on iSCSI, see our iSCSI SAN Buyer's Guide, as well as "State of the Art: How To Build an iSCSI SAN".

Most, but not all, of these devices are based on Windows Storage Server, and vendors have generally added user interfaces to help you manage backups and streaming to disk. Most appliances also run off of S-ATA drives.Although technically inferior to their "enterprise-class" counterparts, S-ATA drives aren't unreliable. For example, the gap between SCSI drives' MTBF (mean time between failures) and that of S-ATA drives isn't as wide as vendors would have you think. Five years ago, the reliability of SCSI was about where S-ATA is today, and SCSI infiltrated the enterprise in short order.

Granted, S-ATA disks are slower, but that's no big deal where backups are concerned. After all, the bottleneck for backups over the LAN is the LAN itself; S-ATA drives can read and write much faster than your network can deliver the data. Remember that drive speeds are in bytes per second, whereas network speeds are in bits per second!

The other piece of this issue is RAID. If you use an appropriate RAID level, you won't lose data in the event a single drive fails (read a RAID tutorial).

Remember, too, that there's a time and place for D2D2T appliances. In a small business or a single enterprise department, they can perform backups admirably, but we wouldn't recommend them for your overall enterprise backup strategy. Backups for a large enterprise require more advanced features, including large tape arrays. Check out the various products available, especially those we've listed in the chart at left, and determine what's right for your organization.

ILM Marketing

One piece of marketspeak that all IT people should be aware of is information life-cycle management. The concept of ILM is that each piece of your data has a value and an optimal storage place: high-end disk, low-end disk and finally tape, depending on the data and the backup/restore needs of the application that uses it.

Now don't get us wrong: ILM is real and can help some organizations meet growing data needs (see "ILM: Panacea or Proprietary Poison?"). Truth is, however, no one vendor or technology can give you an "effective ILM solution."

Most seasoned storage vendors, like EMC, will happily (and truthfully) tell you that ILM is a corporate strategy that can be assisted by technology. But no matter how hard a vendor tries to sell you its magic ILM box, you must resist. Repeat after us: Unless a device is capable of classifying each piece of your corporate data and tracking each classification to its logical end-of-life, it does not, in itself, form an effective ILM strategy.When a vendor slings its marketspeak, ask if the standard contract includes it sending people on-site who will learn your business, create classifications for all your data, inventory your IT assets, determine which types of storage each machine uses and design a beginning to end-of-life strategy for your data given the storage available. If the vendor agrees to this plan (for less than six figures), buy the product and drop us a line. Somehow, though, we expect that the vendor will decline this challenge.

There are plenty of organizations out there that offer "Storage Vaults" -- remote back-ups to their facilities that are managed by the vendor. These services are viable for some organizations but introduce the Internet and your Internet connection as bottle-necks. They also assume that you have Internet connectivity -- something that might not be true, depending on what part of your network needs restoring. While they are cheaper to get into than all-in-one backup appliances, these services also introduce monthly fees -- something that may be a negative to your organization.

QUESTIONS TO ASK PROVIDERS

• How do I recover from complete system failure -- including no network connection?

• How much disk space comes with my plan, and how much do I have to pay if I go over?• What does your organization do to insure the safety and security of my data?