Selecting Backup Media

Selecting a backup medium should always attempt to best meet the requirements of rapid restoration, data integrity and flexibility. The four main media currently in use include tapes, disk drives, Zip and Jaz drives, and CD writing and CD rewriting technologies. Capacity and reliability criteria must also be considered ”for example, while tapes are generally considered very reliable for bulk storage, tape drives are much slower than a hard drive. However, a 20G tape is much cheaper than an equivalent capacity hard drive, so the cost of any backup solutions must be weighed against the value of the data being stored.

Tip  

For more information on choosing a bulk storage device, see the FAQ for the USENET forum comp.arch.storage at http://alumni.caltech.edu/~rdv/comp-arch-storage/FAQ-1.html .

Tape

Solaris supports tape drives from the old Archive Quarter Inch Cartridge QIC 150 1/4 tape drives (with a maximum 250M capacity), up to modern digital audio tape (DAT) and DLT systems. The QIC is a low-end drive that takes a two-reel cassette, which was widely in many early Sun workstations. DAT tapes for DDS-2 drives have a capacity of 4 “8GB, while tapes for the newer DDS-3 standard have 12 “24GB capacity, depending on compression ratios. DDS-2 drives can typically record between 400K and 800K per second, again depending on compression ratios. The transition from analog to digital encoding methods has increased the performance and reliability of tape-based backup methods, and they are still the most commonly used methods today. On the other hand, digital linear tape (DLT) drives are becoming more popular in the enterprise because of their very large storage capacities ”for example, a Compaq 1624 DLT drive can store 35 “70GB, depending on compression, which is much more than the DAT drives. They also feature much higher transfer rates of 1.25 “2.5 megabytes per second. Of course, DLT drives are more expensive than DAT drives, and DAT drives have always been more costly than a QIC, although a QIC is generally much too small to be useful for most systems today.

Hard Drives

Since hard drives have the fastest seek times of all backup media, they are often used to store archives of user files, copied from client drives using an SMB service. In addition, hard drives form the basis of so-called RAID systems, or redundant array of inexpensive disks. Thus, an array of RAID drives can work together as a single, logical storage device, and collectively act as a single storage system that can withstand the loss of one or more of its constituent devices. For example, if a single drive is damaged by a power surge, then depending on the level of RAID protection, your system may be able to continue its functions with a minimum of administrator interference, with no impact on functionality, until the drive is replaced. Many systems now support hot-swapping of drives, so that the faulty drive can be removed and replaced , with the new drive coming seamlessly online. You may be wondering why, in the days of RAID, anybody would consider still using backups : the answer is that entire RAID arrays are just as vulnerable to power surges as a single drive, and so in the event of a full hardware failure, all your data could still be lost unless it is stored safely offsite in a tape or CD-ROM. To circumvent concurrent drive corruption at the end of a disk s life, many administrators use drives of equivalent capacities from different manufacturers, some new and some used, in a RAID array. This ensures that drives are least likely to fail concurrently.

RAID has six levels that are numbered 0 “5, although RAID levels 0 and 1 are most commonly used. RAID level 0 involves parallelizing data transfer between disks ” spreading data across multiple drives, thereby improving overall data transmission rates. This technique is known as striping. However, while RAID level 0 has the ability to write multiple disks concurrently, it does not support redundancy, which is provided with RAID level 1. This level makes an identical copy of a primary disk onto a secondary disk. This kind of mirroring provides complete redundancy: if the primary disk fails, the secondary disk is then able to provide all data contained on the primary disk until the primary disk is replaced. Because striping and mirroring consume large amounts of disk space, they are costly to maintain per megabyte of actual data. Thus, higher RAID levels attempt to use heuristic techniques to provide similar functionality to the lower RAID levels, while reducing the overall cost.

Tip  

RAID level 4 stores parity information on a single drive, which reduces the overall amount of disk space required but is more risky than RAID level 1.

Software RAID solutions typically support both striping and mirroring. This speeds up data writing, and makes provisions for automating the transfer of control from the primary disk to the secondary disk in the event of a primary disk failure. In addition, many software solutions support different RAID levels on different partitions on a disk, which may also be useful in reducing the overall amount of disk space required to safely store data. For example, while users might require access to a fast partition using RAID level 0, there may be another partition that is dedicated to a financial database, which requires mirroring (thus RAID level 1). Sun s DiskSuite product is currently one of the most popular software RAID solutions.

Tip  

Solaris 9 provides integrated RAID support.

Alternatively, custom hardware RAID solutions are also proving popular, because of the minimal administrative overhead involved with installing and configuring such systems. While not exactly plug and play, external RAID arrays such as the StorEdge A1000 have many individual disks that can be used to support both mirroring and striping, with data transfer rates of up to 40 megabytes per second. In addition, banks of fast caching memory (up to 80MB) speeds up disk writes by temporarily storing them in RAM before writing them to one or more disks in the array. This makes the RAID solution not only safe, but significantly faster than a normal disk drive.

Zip/Jaz Disks

Zip and Jaz drives are portable, magnetic storage media that are ideal as a backup medium. Only SCSI interfaces are fully supported under Solaris, although it may be possible to use ATAPI interfaces on Solaris x86. USB and parallel port interfaces are presently unsupported under Solaris. Zip drives come in two storage capacities: the standard 100M drive, and the expanded 250M drive, which is backward compatible with the 100M drive. The 100M and 250M drives are not going to get you very far with backups, but Zip drives do have relatively fast write speeds compared with tape drives. Zip drives are most useful for dumps of database tables, and/or user files that need to be interchanged with PCs and other client systems.

Jaz drives offer several improvements over Zip technology, the most distinguishing characteristic being increased storage capacity. Jaz drives also come in two flavors: the standard 1G drive, and a newer 2G version, which is backward compatible with the standard drive. The Jaz drive is also much faster than the Zip drive, with reported average seek times of around 10ms. This makes Jaz drives comparable in speed to many IDE hard drives, and provides the flexibility of easily sharing data between server and client systems.

Further discussion of Zip and Jaz drives can be found on the alt.iomega.zip.jazz USENET forum.

Caution  

Zip drive technology has improved in recent years , but early versions of the 100M drive suffered from a problem known as the click of death, where a drive would fail to read or write, and a number of repetitive clicks were heard from inside the drive. This problem has now completely disappeared with new models, and users should feel confident in using Zip as a storage medium. For historical information on the click of death problem, see Steve Gibson s page at http://grc.com/clickdeath.htm .

CD-Rs and CD-RWs

CD writing and CD rewriting devices are rapidly gaining momentum as desktop backup systems, which are cheap, fast, and in the case of CD-RW, reusable. CD-R and CD-RW devices serve two distinct purposes in backup systems: while CD-RW disks are very useful for day-to-day backup operations, because they can be reused, CD-R technology is more useful for archiving and auditing purposes. For example, many organizations outsource their development projects to third-party contractors ”in this case, it is useful for both the contractor and the client to have an archival copy of what has been developed in case there is some later disagreement concerning developmental goals and milestones. Alternatively, contracts involved with government organizations may require regular snapshots to satisfy auditing requirements. Since CD-R is a write-once, read-only technology, it is best suited to this purpose. CD-R is wasteful as a normal backup medium, because writable CDs can only be used once. CD-RWs can be rewritten hundreds of times, and with over 600 megabytes of storage, they are competitive with Zip drives for storage, and much cheaper per unit than Jaz drives.