Phil Storrs PC Hardware book

Backing up Data

The need for low cost backup storage systems is becoming more important now that more "mission critical" applications are making their way onto Desk Top and Server PC hardware. Computer System Managers must rely heavily upon sophisticated backup and recovery solutions to ensure data access and integrity.

For desk top systems where we are required to backup the system onto a drive inside the computer, we can use Floppy Disks, Magnetic Tape or High Capacity Removable Magnetic Media . If the computer is on a local area network we can backup over the network to larger drives on a file server or more commonly, to a tape or optical backup system.

Four backup technologies are is use with PC hardware:

Full or incremental backup

We can create an image of the entire hard disk drive, this is called a full backup, but we can save time by just backing up the data that has changed since the last time a backup was done. This is called an incremental backup.

It is common procedure to do a full backup once a week, and to do incremental backups on the other days of the week. If we need to restore the data from backups created this way, we must start by restoring the full backup first, and them restoring each incremental backup in the order they were made in. This means if the system failed on the last day of the cycle, we would have to restore five to seven individual backups.

When working with desk top systems it is often sufficient to just backup the actual data and not the operating system or applications. These can usually be restored easily and quickly from the original disks or CDROMs. To do this efficiently we must setup our directory structures and the applications so the data is stored in easily accessed and identified sub directories, and we must educate the users in how to save data in the correct areas.

Backup software can be made to do an unattended backup overnight or while the user is expected to be out for a break, and this will not disrupt the users productivity. All that is required is for suitable media to be left in the drive. This procedure can also be done over a network and all users data stored on a high capacity storage device in some central location.

Partial backup

It is often economical to backup just the individual data related to one project, application or user, and this can be done, for example, as projects are finalised. The only problem with this approach is to make sure the backup is redone when any changes are made to the data at a later date. High Capacity Removable Magnetic media, CDW and CDRW media are ideal for this purpose.

Making data very safe

Another evolving area in the data storage market that also has a bearing on what type of backup is required is disk array technology, which includes Redundant Array of Inexpensive Disks (RAID). RAID systems are designed to be "fault tolerant" and many allow failed hard drives to be replaced and the data reconstructed without shutting the system down. The process of replacing a drive in this manner is called Hot Swapping.


Storing redundant information which can be used to recreate data when a drive fails, can help prevent data loss. This is the approach used in redundant arrays of inexpensive disks, or RAID systems.

RAID uses multiple disk drives to achieve higher data availability. The redundancy of the disk drives in RAID ensures that no data is lost due to the failure of a single drive unit.

The term RAID (for Redundant Array of Inexpensive Disks), originated at the University of California at Berkeley, in the late 1980s. Most RAID systems use SCSI drives and controllers but some levels of RAID can be implemented using EIDE drives. The actual RAID process in some levels can be performed by the operating system or by the controller.

RAID level 0: Striped disk array without fault tolerance. Data is broken up into blocks and each successive block is written to a separate disk drive in the array. Can be implemented with a minimum of two drives. This is not true RAID because it does not provide Fault Tolerance of any sort and is said to have no redundancy, all the drives in the array are used to store data. If one disk in the array fails the data is lost. Low cost way of achieving high storage capacity from smaller disk drives, fastest access performance is achieved by having one controller per disk drive. RAID Level 0 can be used with cheap EIDE drives and Windows NT to provide greater storage capacity from two three or four smaller hard drives.

RAID Level 1 - Mirroring: Each disk in the array has a corresponding mirror disk and the same data is written to each drive. When data is read back, the data from each drive is read back and compared. This technique provides high reliability as this system is said to have 100 percent redundancy. This doubles the cost of each Mbytes stored. If one disk in the array fails the data on the other disk can be simply copied onto a replacement disk, restoring the data security quickly.

RAID Level 1 - Duplexing: This is a variation of mirroring which duplicates the controllers as well as the disk drives. Both mirroring and duplexing require twice the disk space provided by the array. Mirroring and Duplexing are the simplest RAID systems and are usually done by the Operating System, lowering overall system performance.

RAID Level 2 - Hamming Code ECC: Data is written bit by bit across multiple drives and another set of drives hold (ECC) Error Correction Codes that are used to check the data when it is read back. On read, the ECC codes are used to verify the data and can correct single word disk errors on the fly. With two or more ECC drives in the array, data can be reconstituted onto a replacement drive when one drive in the array fails. Extremely high data transfer rates can be obtained, and hot swapping of failed drives is usually possible.

RAID Level 3 - Parallel transfer with parity: The array has multiple data disk drives and only one parity drive. The data is stripped (written bit by bit) across the data drives. The controller can determine which drive in the array has failed using additional check-sum information recorded at the end of each sector on the data drives. Although the total number of disks required for a RAID Level 3 subsystem is lower than that of a RAID Level 2, the usable capacity of the disks is less because of the extra check information required on each sector on the data disks. The RAID Level 3 technique is usually provided by the controller card as it is difficult to do with software. It provides very high data transfer rates and a disk failure will have low impact on the system throughput.

RAID Level 4 - Independent Data disks with shared Parity disk: Similar to RAID Level 3 in that it requires multiple data drives and only one parity drive but each whole block (sector) of data is written on the same drive. In RAID Level 4 data is written across the drives sector by sector. The controller design is complex and rebuilding a failed drive is slow.

RAID Level 5 - Independent Data Disks with distributed parity blocks: Each data block is written on same drive but parity for each block is distributed across the disks in the array. Good overall data transfer rate, complex controller design, difficult to rebuild a drive if one fails.

RAID level 6 - Independent Data disks with two independent distributed parity schemes: An expansion of RAID Level 5 that provides an extra level of fault tolerance by using a second independent distributed parity scheme. The best solution for "Mission Critical" applications but requires a very complex controller and has poor write performance.

RAID level 7 - Optimised Asynchrony for high I/O rates as well as high data transfer rates: This is a proprietary RAID system from IBM that is very expensive but provides the fastest overall performance.

RAID level 10 - Very high reliability with high performance: This is a hybrid implementation which uses features of both RAID-0 and RAID-1. Block striping of data is done at the operating system level and parallel mirroring is done at the disk controller level. It is very expensive to implement, has the same overheads and fault tolerance as Mirroring but provides higher data throughput.

RAID level 53 - High I/O rates and data transfer performance: This should be called RAID 03 because it is implemented as a striped (RAID Level 0) array whose segments are RAID Level 3 arrays. Expensive to implement but does provide high data transfer rates with RAID Level 3 fault tolerance.

Exotic hardware

In very large Server installations, Juke Box Disk or Tape changers are used to swap media in and out of Tape or Optical drives as required. This technology can provide vast automated storage space for data backup or archival storage. As many Mainframe computer systems are replaced with powerful PC based hardware, the need for keeping vast amounts of data very safe and with a very high availability factors is now very important. RAID arrays combined with hot swapping technology and fast reliable backup achieves these aims.

Back to the opening index Book four index

Copyright © Phil. Storr, last updated 4th December 1998