Tape turning: protect against data loss.

Using disk to replace tape for backup and disaster recovery operations is one of the big stories in the data storage market this year. On the surface, the idea makes sense: disk is well-understood, fast, reliable and inexpensive--or at least that's what the VTL (Virtual Tape Library) and D2D (Disk-To-Disk) marketers promise. But many users have taken a look at disk-based data protection options and come to the conclusion that they aren't ready for prime time, at least not yet. They seek to improve the costs and efficiency of their data protection systems, but some issues are keeping them from doing so:

* Disk hardware costs more than both tape hardware and tape media, a lot more. Even low-priced ATA-based drives are several times the cost of tape, on a capacity basis.

* Disk-based solutions do not provide an easy, cost-effective way to store a copy of your data off-site; you either need to replicate the data to another disk-based system in another location (over expensive transmission lines) or use a local tape system to enable off-site transport and storage.

* Add in the electricity and cooling costs of all those spinning disks, plus the costs of scaling to higher capacities, and you're looking at a significant investment; and the pay-back is only seen in somewhat faster recovery times.

* Some users have problems with volume labeling compatibility when they use both disk and tape.

It seems strange to some to use "cheap" disk to protect the data on "expensive" disk. If the ATA disks in the VTL or D2D systems are reliable enough for a data protection solution, why aren't their high-end SCSI and FibreChannel cousins reliable enough to avoid the need for backup in the first place? Actually, by adding more magnetic disk into the system, aren't you just adding more opportunity for data loss?

"While disk has received a great deal of attention lately as a backup-and-restore media, it should not be viewed as an all-out replacement for tape," says Dianne McAdam. Senior Analyst and Partner of Data Mobility Group. "Rather, it should be viewed as an addition to existing backup processes. In fact, D2D should be considered as part of a larger backup strategy of primary disk to secondary disk to tape, or D2D2T. While disk can be used to store the most current versions of backups, tape continues to maintain its place in the data center for retention of long-term backups and archival storage. And tape storage costs much less than the equivalent amount of disk storage. Not only does disk storage initially cost more than tape, but continues to costs more throughout its life when environmental factors such as power, cooling and floor space costs are added to the equation."

So, if you think that a disk-based data protection solution might not be right for your company, the next best thing is to improve your tape operations to reduce or even eliminate the problems that may have gotten you to think about a disk-based solution in the first place. The hearsay issues with tape include reliability, performance, and utilization. Let's take a look at approaches to improve each of them.

Reliability

Today's professional tape systems are designed to provide robust and reliable data storage for many years. Even so, there are several factors that feed into the perception that tape is inherently unreliable. As with anything, one of the primary factors is human error: most data recovery failures are due to a failure to perform the backup in the first place.

"Automate data protection programs as much as possible to eliminate the
chances for human error. Document the process and periodically review it
for opportunities to improve."
--Dawn Wortman, Fujifilm

The second most-common failure factor is from improperly handling tape media. When transported in its original packaging, and stored in a controlled office environment, professional-grade tape will reliably retain data for at least 15 years--far longer than almost any data protection or archival application requirement.

"Follow the manufacturer's recommendations for tape handling and
storage. Excessive heat or humidity will reduce the life and reliability
of magnetic tape."
--Jim Milligan, Imation

Data errors and other problems can occur if the tape drive is not maintained correctly. The most important aspect of maintenance is periodic cleaning of the read/write heads. Many drive formats use a special cleaning cartridge to perform this operation automatically, and some drives even prevent writing new data if they have gone too long without a cleaning cycle.

"Ensure that cleaning cartridges are only used for their recommended
number of cycles and that other maintenance schedules are followed, per
manufacturer's recommendation."
--Paul Scheuer, IBM

A growing source of tape problems is due to the use of recycled tape cartridges. As with buying any used products, you may be purchasing someone else's problems and in the end, will not be saving money by using recycled media. You never know how these tapes have been used, handled and transported, or even what source of media may be in the cartridge.

"Only purchase new name-brand tape media from reputable resellers;
ensure that the cartridge comes in its original protective packaging."
--Steven Potcher, Maxell

Remember that you're performing a backup because of potential data loss problems arising from the use of magnetic disk. Some companies that rely on disk for data protection keep as many as 25 copies of critical data in various locations to protect against any and all risks (talk about storage management headaches!). But once the data makes it to tape, the prevailing belief is that it's now in safe hands.

Performance

One of the big misconceptions about tape is that it's too slow for today's business environments. In fact, current version mid-range and enterprise-class tape drives can read and write data at extremely fast rates, ranging between 50 and 75MB/sec when using data compression. This is faster than most networks can feed a tape drive, and therein lies the problem.

As a streaming technology, tape drives work best when data is coming to them at their rated speed. Unfortunately, many networks are not able to provide data that quickly, due to other network traffic, latency, transfer protocol overheads, etc. And this leaves the drive waiting for its buffer to fill before it begins writing. Once it sends its buffer to the tape, it then has to stop, move back a little, and wait until the buffer fills up again. The resulting stop-reposition-start cycling of the tape during write operations is sometimes referred to as "shoe-shining", and this can seriously cut the effective performance of a tape drive and result in longer backup times.

The use of a centralized backup server or other caching mechanism can mitigate this variable, as the transfers to tape will occur directly between the server/cache and the tape drive in the background. Some library manufacturers are now including a caching layer within their libraries. Another common approach, supported by several major backup software applications, is to interleave the backup data from several application servers into a single stream for the tape drive.

"Ensure that your backup application and infrastructure can feed data to
the tape drive at least as fast as the drive's rating. If not, the drive
can be slowed down by turning off compression (at the expense of
capacity). Consider using a backup server or front-end cache."
--John Goode, Quantum

Dirty tape heads or problems with the tape itself can result in read/write errors. Given the sophisticated error recovery procedures built into tape drives, performance may be affected by repeated attempts to write or read the data correctly; data integrity takes precedence over speed. These errors and recoveries will show up in the administrator logs, but to the user, the system will just seem very slow.

"Maintain the drive per the manufacturer's recommendation, and ensure
the tapes that are in use to be authentic and approved/certified for use
in your drive."
--Chris Smith, Sony

The data transfer rate of a tape drive is directly related to its capacity. When the drive and media technologies go through periodic upgrades, usually involving a doubling of capacity (i.e. from 100GB to 200GB), the data transfer rate of the tape system is increased proportionally (from 12 to 24MB/second).

"If you need increased performance, it may be worth looking at a newer
generation of the tape technology you currently use, or maybe consider
switching drive types. Some library manufacturers support multiple drive
types, so different drives can be used for different tasks or
applications within a single library."
--Bob Raymond, StorageTek

Tape performance is most often a function of how it's used, rather than one of its design specifications. If you can take advantage of the data transfer speed that tape offers, then the costs of adding a disk-based solution will seem even less justifiable.

Utilization

The good news about the improvements being made in tape technology is that the capacity of tape cartridges continues to double every 2 to 3 years. With the MSRP of the tape drives, media, libraries and software remaining constant in dollar terms, the cost of data protection is effectively cut in half with each subsequent upgrade cycle.

The bad news, if you can call it bad, is that these larger tape capacities may be under utilized. While a full tape may be needed to perform a full backup, significantly less may be needed for daily incremental backups. As much as 90% of the tape capacity may be wasted--at least until it is passed through the backup retention cycle and re-introduced for erasure and re-use.

"A single tape cartridge can be used to record multiple backup sessions,
thereby increasing its utilization. It may also be possible, depending
on your system architecture, to store incremental backups from multiple
applications on a single tape cartridge."
--Henry Dobashi, TDK

Tape systems and applications offer many options for getting the most value from your data protection investments. You can refer to your vendor for other ideas specific to their product or contact the Tape Technology Council, www.tapecouncil.org.

Summary

The arguments made for disk-based data protection systems are interesting, but on further examination, their benefits do not match the promises. In reality, adopting a disk-based backup system may provide a small performance improvement, at a higher cost than tape, all without eliminating the need for tape or some other disaster recovery solution. Before going down that path, it would be prudent to take a look at what you may want your current data protection system to do and try to improve it with the current technology being used. Your system integrator and service provider have knowledge about the tuning and configuration of your system to best meet your operational requirements.

Tape solutions continue to be the most cost effective and provide the best data protection, back up and archiving functionality. To that end, we don't see products that aim to backup and protect data that resides on tape, other than to make copies for off-site disaster recovery purposes, yet there are hundreds (thousands?) of products available to protect data that's on disk.

Rich Harada is president of the Tape Technology Council.

www.fujifilm.com

www.ibm.com

www.imation.com

www.maxell.com

www.quantum.com

www.sony.com

www.storagetek.com

www.TDK.com

www.tapecouncil.org

Opening shots in continuing stories ...