Tape archive technology, Part I

You’ve heard the statement, “Tape is dead.” Let me correct the statement: Not by a long shot.

The choice of backing up content on disk or tape, Linear Tape Open (LTO), sometimes gets video folks all heated. If the video engineers propose that all video content be stored on real-time disk, the IT department may at first push back saying, “Let’s use tape. It’s cheaper.” The media folks may respond, “But it's unreliable and slow.” The truth is some of both opinions.

What is the solution to video for backup or disaster recovery purposes in a production environment? Should content be stored long-term exclusively on rapid-access disk, or on lower-cost tape? What about the cloud? The best solution is more complex than it may first seem.

Gone are the days when all content was stored on tape. Today, tape may be used at ingest or even during high-end production, but most content in production environments is stored and edited from disk. So, the question about using tape for anything seems obtuse.

The arguments for one technology over the other merely confuse the issue. On one hand, a disk-centric storage pool is fast and provides random access. Disk storage also is reliable and easily expandable.

Tape, on the other hand, is often perceived as a slow, unreliable and linear process. In other words, tape is often considered old-fashioned. However, from a cost perspective, a tape-based solution is less expensive and new LTO technologies make the technology worthy of consideration, especially when there is the need for a large storage silo.

Often times, the best solution is a blend of disk and tape technology. To more closely examine the options, we’ll begin this journey about storage by examining new LTO technology. You’ll learn how data is recorded on linear tape and some factors to consider when choosing a production backup technology.

TCO

Any discussion about buying a new storage technology involves more than just CAPEX. Ongoing expenses (OPEX) must be considered also. An IT manager or video engineer will need to determine not only what the initial purchase costs will be, but also what the ongoing supply and maintenance costs may total.

Let us compare a conventional disk-based, RAID array for media storage with LTO-5 tape technology.

For tape technology lovers, this next chart will be all you want to know. Tape wins hands down when it comes to the cost of long-term data storage. In one study presented by Bruce Master, IBM sr. manager, LTO program, the cost of spinning disk storage is shown in every case to be higher than for tape. The results of his study are summarized in Figure 1 to the right. Figure 1 shows the OPEX cost comparison between tape and disk storage platforms. While this data shows tape storage to be less expensive than disk storage, this alone should not be the singular purchasing decision factor.

The study shows that:

• The cost of storing a single TB of data long-term on a spinning disk is 23 times higher than for tape.

• The electrical cost to operate a disk system is 290 times higher than for tape.

• Even when combined with a 20X data deduped virtual tape library, the OPEX of disk may still be five times more costly than tape.

Under the conditions set forth in the study, when measured over time, tape is the clear winner in the cost category. If you have to justify a backup system solely on OPEX, tape would appear to be the best option.

There are other factors that must be considered when selecting a media storage solution. Before we get to some of those issues, let’s look inside the technology behind LTO systems.

Technology behind LTO

LTO storage technology is, in some ways, similar to what is (was) used in VTRs. While spinning heads, like those on a VTR, are not used, the tape is shuttled back and forth along a moving head assembly creating multiple tracks. Let’s look at some of LTO’s basic technology. For this discussion we’ll be looking only at LTO-5 technology.

LTO technology is available in two general formats: Accelis and Ultrium. Both formats use the same technology for:

•heads

•media track layout

•channel technology

•servo technology

They also share other common electronic building blocks and some common code blocks. The Ultrium format uses a single-reel tape cartridge and a wider tape permitting high data capacity. The Accelis format uses a dual-reel cartridge and a midpoint load to provide shorter data access time.

A note about versions of LTO tape decks. LTO-5 is the most current technology. It is backward read compatible with LTO-3 and LTO-4 decks. LTO-3 and LTO-4 decks cannot record LTO-5 formatted data. So, if your facility has those version decks, they will not be forward compatible.

LTO-5 storage achieves its high density by recording six tracks of eight channels of data. It does this through a clever servo mechanism, which positions the head assembly using five prestriped six-position servo tracks spaced across the width of the data tape.

A cross section of an LTO-5 tape is shown in Figure 2 to the left. (The Figure shows an LTO-5 tape is divided into bands: four data bands, five servo bands, each with six prerecorded tracks and two guard bands. Each data band contains six rows of eight-channel data written sequentially forward and backward on the tape by a 16-element read/write head plus four servo head assembly.) An LTO tape, much like videotape, has multiple bands for both data and control signals. An LTO-5 tape has four bands of data, numbered zero to three. Edge-located guard bands are located at both the top and bottom of the tape to protect the servo and data bands. The write/read head assembly spans one data band and writes sequentially eight rows of data in each direction of tape movement. Servo bands, each with six tracks, are positioned at the top and bottom of each data band and used to position the write/read heads. The servo bands are written onto the tape at the time it is manufactured.

Let’s look more closely at how the data is recorded. The write/read head is controlled by a servo system that uses information prerecorded in five servo bands across the tape. Two servo bands are used simultaneously to provide dual references for servo positioning. (See Figure 3 to the right, showing the striping of the servo bands. Four servo heads, used in pairs over separate servo bands, accurately position the data head.)

Each servo band contains six tracks, which are read simultaneously for precise positioning by two of the four servo heads. During a write operation the top servo zero, or TS0, is centered on servo position zero, and the eight write elements write data in the first track group. When the head senses the end of the tape marker, top servo one (TS1) takes control and aligns on servo position five, and the second group of data are written in the reverse direction. A total of 384 tracks of data are written across the tape as it shuttles back and forth, end-to-end. An animation of the record process can be seen online.

Data is first written and then immediately read to ensure an accurate recording. If an error occurs, the same data is rerecorded farther down the data tape. It takes only a few minutes to write one set of data tracks along the entire length of the tape. An in-depth (500-plus page) tutorial is available.

Reliability

With this brief LTO background, let’s now examine how LTO recording compares to disk storage in other ways. First reliability.

Disk storage comes complete with its own built-in data protection. It’s call RAID. Tape has no such protection. Gartner Research and Storage magazine report that 34 percent of companies who used tape as backup never tested their backup systems. Of those companies that did test their backed up data, 77 percent discovered backup errors. Some firms attempt to address this shortcoming by creating multiple copies of backup tapes. However, this merely compounds the problem and greatly increases data management issues. Let’s see how a typical backup process can greatly multiply the number of tapes needed to help ensure reliable data recovery.

One backup process is called a Grandfather-Father-Son (GFS) managed retention. In this model, daily incremental backups are made combined with full weekly backups. While this sounds efficient enough, the result is that over one-year’s time, for every TB of data, fully 25TB needs to be written to tape for protection.

Figure 4 to the left illustrates how the GFS backup procedure may result in a high number of tapes being required. The chart shows that in order to back up 42TB of yearly data using a GFS daily/weekly backup model, it would require 6300 LTO-2 tapes. The cost of just those cartridges at $26 each would be $163,000. The selection of other LTO technologies will affect this total cost.

This amount of typical business data in a deduped format would equal about 2TB. Today, a 2TB drive costs $300. Consider the ease with which that 2TB of data could be quickly moved to an off-site location for disaster recovery.

However, lest you think that disk is obviously the best solution for storage, stop. There are plenty of good reasons to use tape storage in a carefully planned media backup archive. We will continue this discussion next time by looking further at the benefits offered by LTO technology.

CATEGORIES