RAID is the most common method of data protection today for SMBs as well as larger enterprises. Most companies rely on the redundancy provided by RAID at various levels to protect them from disk drive failures. However, RAID is evolving as larger capacity drives and the use of cheaper and less reliable drives become more popular. In addition, some companies are beginning to look beyond RAID and offer proprietary products that take a...
holistic rather than a component-level view of storage and offer increasingly sophisticated tools for managing storage at a finer level of granularity than the disk.
RAID 10 combines mirroring the data to two sets of disks for redundancy with striping the data across disks and within each disk set for better performance. It is conceptually simple and gives good performance, but it requires twice the space to store a given amount of data.
RAID 5 stripes the data across the disk set with added parity blocks for redundancy. In the event of a disk failure, the information on the failed disk can be rebuilt from the parity data. RAID 5 is more economical than RAID 10, requiring only the equivalent of one extra disk per disk set (the parity information is distributed across the entire disk set).
RAID 5 loses performance on write because the system has to calculate and write the parity information. That can amount to as many as four write operations per write to the disk array. It also takes time to reconstruct the damaged disk in the event of a failure.
A combination of increasing disk sizes and greater use of less-expensive SATA and SAS disks has made RAID 6 more popular. RAID 6 uses striping like RAID 5, but it has two parity stripes instead of one, which protects data against two disk failures.
As disk capacities increase, it takes longer to rebuild the RAID array after a disk failure. In fact, in the amount of time it takes to rebuild the RAID array in RAID 5, the possibility of a second failure before the array is rebuilt increases. According to Xiotech, a rebuild of a 1 TB drive in a five-disk RAID 5 array will fail 40% of the time under common conditions (assuming disks with an unrecoverable read error rate of one in 10 to the fourteenth bits).
The future of RAID
Meanwhile, some companies are moving beyond RAID to other methods of insuring disk reliability. Typically, these methods combine RAID features with proprietary methods of monitoring, managing, and repairing disks. Some companies such as Atrato and Xiotech offer sealed arrays that are guaranteed up to three years (Atrato) or five years (Xiotech).
Atrato offers a Self maintaining Array of Independent Disks (SAID) in its Velocity 1000 product. This consists of 160 2.5-inch disks in a high-density 3U form factor enclosure, capable of supporting multiple independent data streams. Diagnostic and maintenance software is designed to automatically detect and correct problems, swapping in spare drives as needed.
Xiotech offers a technology dubbed Intelligent Storage Elements (ISE), which consists of dual power, cooling and battery modules, supporting one or two sealed DataPacs containing up to 8 TB of disk storage per ISE. A module controller is also included as a part of the ISE.
Both products conform to the ANSI T10-DIF standard for end-to-end data protection and detection and prevention of silent data errors. ANSI T10-DIF provides a logical block guard to compare the data actually written to the disk with what is supposed to be written, a logical block application tag to ensure that the data is written to the correct logical unit and a logical block reference tag to see that the data is written to the proper virtual block.
Both Atrato's and Xiotech's products use controllers that closely monitor the health of the disks and can take preventive and corrective action in the case of potential failure. Atrato's controller performs constant low-level background scrubbing and health checks on unused storage to verify it is usable if needed. In the event of a failure, Xiotech's controller swaps the data to unused disk and then uses what it calls "self-healing technology" to recondition the drive. The controller resets or power cycles the drive, recalibrates the heads, rewrites the servo tracks and performs a low-level format. The controllers use sector- or region-level remapping to block out bad sectors and automatically transfer the data to good sectors elsewhere in the array.
RAID is not by any means dying, but the increasing demands of new generations of disks and new generations of storage are opening the way for new approaches that extend disk protection beyond RAID. RAID is likely to remain an important part of data protection, but it will probably be supplemented by additional technologies aimed at making drive arrays more failure-protected and more modular.
About the author: Rick Cook specializes in writing about issues related to storage and storage management.