RAID: An Overview

RAID Versions: An Overview

Selecting the appropriate RAID level requires thorough consideration. Key factors in the decision process are availability, performance and cost per MByte. Organizing the harddrives through JBOD creates the smallest costs, but is not a good option in terms of speed reliability. In contrast, a level-1 array guarantees the highest degree of availability. On the downside, it generates the highest capacity overhead and thus the highest relative costs. If these factors are combined in a diagram, the result is the typical RAID triangle, as shown in the image below.

Behaviour in Case of Harddrive Failure

No matter what the scenario - whether it is harddrive compound within a computer or an external memory sub-system, or a hardware RAID or software RAID: If a harddrive of the array fails, redundancy gets lost in most of the commonly employed RAID levels. Every failure of another harddrive ultimately leads to data loss. The consequence: The damaged harddrive has to be replaced as quickly as possible and the array has to be reconstructed.

Ideally, the array possesses an additional harddrive that is only used in case of an emergency. Such a hot-fix harddrive - also known as hot-spare or stand-by harddrive - is activated automatically and implemented as a substitute for the damaged harddrive. If no hot-spare is available, the damaged harddrive has to be replaced manually. Typically, this requires users to shut down the computer, which, in turn means an interruption of operations. But especially for servers, such a shutdown is unacceptable. The solution: Arrays which are capable of hot-plug and hot-swap. The RAID compound's harddrives are stored in harddrive shuttles, where they are freely available and can be changed, even if they are operating at the time.

After the exchange process, the new harddrive has to be integrated into the RAID compound, and it has to reconstruct the loss data. If this process happens automatically, it is called auto-rebuild. The automatic reconstruction assumes, though, that the controller can communicate with the harddrive shuttle (status of the harddrive, new shuttle was introduced). If controller and shuttle cannot communicate, the rebuild process has to be initated manually.

Conclusion: Redundancy With Limitations

The use of disk arrays significantly increases the availability of computer systems. But at the same time, RAID is no miraculous tool preventing data loss. In order to obtain a 100-percent reliability, all components of the memory sub-system - including controller, power supply and fan - have to be designed in redundant form. While the storage industry offers such solutions, they are not exactly cost-efficient.

In addition, failures of harddrives and other components not always happen independently from one another. In day-to-day operations, there occasionally are situations that significantly increase the failure probability of the entire array. Examples include excess voltage, for instance, due to lightning damage, but also flooding or fires. Also, viruses and worms attack RAID systems just as often as individual harddrives.

After all, even the most reliable array cannot eliminate the No. 1 risk factor - the human. The largest portion of irreparable data losses are not caused by failing technology but are the result of faulty use. Deleted data are lost on RAID systems, too. So, even for the most sophisticated RAID system, the rule of thumb has to be: The only truly reliable protection against data loss is a consistently planned and executed data backup.