Raid Technology - Term Paper Example

Definition: RAID, it is a term which means Redundant Array of Independent Disks that was changed from the initial term which was Redundant Array of Inexpensive Disks (Brinkmann, et al, 2007); it is technology which provides increased storage functions and reliability through redundancy. Features: It has redundant architectures so as to enable the storage facility has the fault tolerance property. Has a higher throughput levels than a single hard disk or group of independent hard drives. The disks used in array are independent of each other thus reducing the cost of implementing the arrays. (Dawkins and Arnold 2006). Architecture: The following are some of commonly used RAID levels: RAID 0: it is the level without parity or zero redundancy. It has a minimum number of two drives and provides improved performance and additional storage but has no fault tolerance and thus the simple stripe sets are referred to as RAID 0. RAID 0 does not implement error checking mechanism and hence any error is uncorrectable (Hennessy and Patterson, 2009). When more disks are used in the array it implies that it has higher bandwidth, but greater risk of losing data. RAID 1: This is the level of mirroring without parity or striping, here data is written identically to multiple disks. Many implementations create sets of two disks but some sets may contain three or more disks. The array provides fault tolerance from disk failures and it functions when at least one of drives is functioning. It has the key weakness which is high cost overhead; this is because data is duplicated as required by storage capacity. RAID 2: it is the level with bit striping that has dedicated Hamming-code parity, in this level there is a synchronized rotation of disks and data is striped such that each sequential bit is on it is own disk (Hennessy and Patterson, 2009). When Hamming-code parity is calculated across the subsequent bits on disks it is then stored on the parity disks available. RAID 3: it is a byte-level striping that has dedicated parity, disks spindle rotate in a synchronized manner and data is striped in such a way that all the subsequent bytes are stored on different disks. Dedicated parity disk is used to store the parity after it has been calculated across the corresponding bytes on disks. RAID 4: it a striping block level that has dedicated parity, in this architecture all parity data is confined to a single disk which can lower its performance. In this arrangement, there is a spread of files between many disks. The disks in this array have independent operations allowing I/O requests to be done in a similar way, though the drawback is that data transfer speeds suffer as a result of the type of parity being used. RAID 5: it is a level of block striping whose parity is distributed. Data is distributed along with the parity and drives are required to be present in order for it to operate implying that if a fails occurs then it will have to be replaced, however, an array is not destroyed when a single drive fails. When the drive fails, any next reads is calculated from the distributed parity such that when drive fails then it is protected from the end user (Hennessy and Patterson, 2009). Design Issues: The booting process is the key concern in the RAID based operating systems. At times it can be difficult to come up with a boot process that can fail to another drive if the normal boot process fails. When the system fails then a manual intervention has to use so as to make the system become bootable after another failure. A hardware RAID controller has an explicit programming which identifies the broken disk. Hardware RAID controllers at times has battery that is covered by cache memory. In the modern system, for data to be safe then the user of software RAID has to ensure that the write-back cache on the disk is off. When the write cache is turned off, it has a performance penalty accompanied with it which relies on the amount of workload and the command queuing in the disk system. The use of RAID controller that has battery backed cache is one of the solutions that are required to have safe write-back cache. Operating system-based RAID uses formats specific to the operating system at hand such that it cannot be used for partitions that are shared between the operating systems in the event of multi-boot system. Dawkins and Arnold (2006) explain that, RAID disk is allowed to move from computer to computer with a file or an operating system of the same type, this is at times is more difficult when using hardware RAID. Challenges: 1. Associated failures: The error corrections in the RAID systems assume that the failures of drives are independent. With this in mind it is possible to calculate how often the system can fail and to arrange the array so as to make data loss arbitrarily importable. In practice, these drives are frequently manufactured at the same time, having similar wear and same environmental exposure. Given that many drives fail as a result of mechanical issues which mostly occur in older drives, then this contradicts the assumption of correlation. 2. Write cache reliability: The disk system normally acknowledges the write operation when data is in the in the cache, implying that it does not have to wait for data to be physically written. This system is applied in old, non-journaled systems such as FAT32 which do not put into consideration the software updates protection (Wilkes, 1996). A power outage or hang in the system such as BSCOD means that there will be a significant loss of any data in the queue in that cache. Sometimes the battery solves this of write cache this is because when power failure makes the write process to fail, then controller completes the pending writes as soon as it restarted. However, Wilkes (1996) explains that, this solution has potential failure cases; the controller may have been worn out and thus the power may be off too long and moving the disk to another controller does not solve the problem because the controller itself could fail. There are some disk systems which have the capability of testing the battery 3. Data recovery: As noted by Cortes (2003), since the disks used are large in size, the odds of disk failure during the rebuild are not small so in with this regard, the process of extracting data with difficulty from a failed state has to be put into considerations. It is only RAID 1 that stores all data on the each disk. Even if the data recovery depends on the controller, the RAID 1 disks are read as single convectional disk implying that a dropped RAID 1 disk while damaged can easily be restored by the use of software recovery program. In the case of severe damages then data is usually recovered by professionally data recovery specialists. RAID 5 and other stripped arrays present much more difficult obstacle to data recovery when the array fails. Operator skills, correct operation So as to come up with desired protection which resist physical drive failure, then the RAID array has to be properly set and maintained by an operator who has sufficient knowledge of the chosen RAID configuration(Dawkins and Arnold. 2006). When unskilled handling is applied on the array at any stage it may worsen the consequences the array failing, resulting in the loss of data which was likely to be recovered. In general, the array has to be regularly checked so as to identify the failures. When the problem is not dealt with, it will result in the array continuing to operate in degraded state and thus being subjected to further failures. This may make the entire array to become fatal, making data to get lost. Performance Software-based RAID: They are primarily used with entry-level servers and relay on standard host adapter. They are used in executing the I/O commands and mathematically intensive RAID algorithm which are hosted in the server CPU. Advantages: The biggest benefit of software based RAID system is that it only requires a standard controller to operate them. The second benefit is its low price. Hardware-based RAID: The controllers are plug into the bus slot, some or all of the I/O commands are offloaded and all or some of the I/O commands are offloaded by the RAID. Advantage: The array has a more resistant fault-tolerant features and an increased performance as compared to software-based RAID. Scheduling: RAID technology came into being so to find solutions of fitting the company’s storage needs. There are very many solutions but the suitable ways of implementing the RAID are picked. If you choose software RAID the cheaper the performance and vice versa (Dawkins and Arnold. 2006). Examples: There are three types of RAID systems which include: a) Failure-resistant disk system: this protects data against failure when the disk fails. b) Failure-tolerant disk system: it protects against the data loss due to failure of any single component. c) Disaster-tolerant disk systems; it consists of two or independent zones, either of these zones provides access to stored data. Conclusion As already discussed, Redundant Array of Independent Disks (RAID) is technology which provides increased storage functions and reliability through redundancy. However, it has its design issues that have to be overcome to use it. References: ______ (2008) Raid Technology: Available from: http://www.redhat.com/docs/manuals/linux/RHL-6.2Manual/ref-guide/ch-raid.html Brinkmann, A et al (2007): Dynamic and Redundant Data Placement, Proceedings of the 27th IEEE International Conference on Distributed Computing Systems (ICDCS) Brinkmann, A, Salzwedel, K and Scheidele C (2002): Efficient, Distributed Data Placement Strategies for Storage Area Networks; Proceedings of the 14th ACM Symposium on Parallel Algorithms and Architectures (SPAA), pages 53 - 62, Cortes, T (2003): Taking advantage of heterogeneity in disk arrays, Journal of Parallel and Distributed Computing, Volume 63, Issue 4, Pages: 448 - 464, Dawkins, B and Arnold, J (2006): Common RAID Disk Data Format Specification, Colorado Springs. Eliezer, L & Silberschatz A: (1990): Distributed file systems: concepts and examples, ACM Computing Surveys (CSUR): Volume 22, Issue 4, pages 321 - 374, Hennessy, J. and Patterson, D. (2009): Computer Organization and Design: Morgan Kaufmann Publishers, New York. Wilkes, J (1996): The HP AutoRAID hierarchical storage system, ACM Transactions on Computer Systems (TOCS); Volume 14, Issue 1, 108 – 136. Read More