|
|
|
|
RAID--Redundant Array of Independent Disks
A RAID controller uses a variety of hardware and/or software configurations to store and protect data. If you need to safeguard your company's information from hardware and/or software failures, read on to learn how RAID storage can benefit you.
Unlike a typical hard drive, a RAID array (a group of two or more disks) can provide a higher level of data security, data integrity, fault tolerance, data redundancy and in many cases, increased performance. Each RAID level provides different features, functions, tradeoffs and benefits. Assess your needs--do you need to protect mission-critical data? How long can your network be down in the event of a failure? Do you need "off site" storage? Are you looking for the highest performance with the least down time? Are you looking for devices you can hot-swap?
RAID technology is a great way to increase the fault tolerance and reliability of your server environment. If you have applications that are mission critical to your business, one of the many available RAID solutions offered by Bloom MicroTech will give you the peace of mind you expect with your new system. RAID solutions are not meant to be a substitute for a good backup system. Backup units like tape systems and other removable media allow you to take snapshots of your data and go back in time. Backup systems also allow you to take your snapshots off site in the event of a natural disaster or unforeseen emergency. This means no matter what happens, you can start where you left off.
This overview will help you determine which RAID design will meet the needs of your system. Further information on configuring Eclipse servers with RAID solutions is available through Bloom MicroTech's Network Solutions Team.
NOTE: The number of the RAID level does not indicate the level of fault tolerance. Instead, the RAID level only designates the method of RAID.
RAID-0--Striped Disk Array without Fault Tolerance--2 drives minimum
RAID-1--Mirroring--2 drives minimum
RAID-2--Hamming Coded Disk Array--not used today
RAID-3--Byte Striping with Single Drive Parity--3 drives minimum
RAID-4--Segment Striping with Single Drive Parity--3 drives minimum
RAID-5--Segment Striping with Distributed Drive Parity--3 drives minimum
Striped Disk Array without
Fault Tolerance
RAID-0 stores data across multiple disk drives in a configuration called a stripe set. In a stripe set, data is distributed evenly over drives on identically-sized areas of the disks. The stripe set in the array of disks performs as one large, fast drive.
There is no wasted space in a RAID-0 configuration. 100% of your disk capacity is available for the storage of data. All disk drives work together at the same time as though they were a single drive. The increase in access speed is attributed to the fact that the I/O load is spread evenly across all drives in the array. There is no overhead for a parity calculation in a RAID-0 configuration. This combined through-put and low overhead helps to give you the fastest performance of all the RAID configurations.
It is important to note that RAID-0 does not specify any fault tolerance. If a single drive fails, all data in the RAID array will be lost. For this reason, you should never implement RAID-0 by itself for your mission-critical data without a reliable backup system in place.
Applications that require very high-speed storage but do not need immediate fail over capability would be well suited to a RAID-0 type configuration. RAID-0 works well for applications that create large temporary files, streaming video, streaming audio, video-on-demand, and applications that need to manipulate large graphic file types very quickly.
Mirroring
A RAID-1 configuration uses mirroring to duplicate block for block the data written to your source or working drive. Each disk in a RAID-1 configuration will have an identical backup. These pairs of disks are called "mirrored sets" or "mirrored pairs." They operate as though they are one drive of the same size and type. If one drive should fail, you will have a complete working backup of your mission-critical data. Mirroring provides full fail over capability so there is no interruption of service. Mirroring is done in real time with no additional overhead or impact to performance. Duplexing is another technology utilized with RAID-1 that uses a separate disk channel for each drive in a pair.
Replacing a failed drive in a RAID-1 configuration is as easy as installing the new drive and telling the two disks they are now a "mirrored pair." The two drives will work together as one and any data that needs to be copied to the new drive will be done automatically.
RAID-1 is the most reliable of all RAID configurations, but it does cost more to implement. You must double your investment in disk storage to get the benefits of online backups. If your enterprise relies on complete, accurate and secure data storage, a RAID-1 design may provide the best fail over solution and the peace-of-mind you need.
Hamming Coded Disk Array
RAID-2 is similar to RAID-0 because it stripes data across two or more disks in an array. It uses Hamming error correction codes to protect data on disks which do not have built-in error checking and correction (ECC). ECC is the most reliable means of ensuring that data is written to a drive correctly. RAID-2 is not generally used in SCSI systems today because SCSI hard drives automatically perform their own error checking and correction.
Byte Striping with Single Drive
Parity
RAID-3 combines data striping across two or more disks with a single dedicated parity drive for data protection. The parity disk can reconstruct data on a failed data drive by comparing information to the remaining data disk(s).
Because the majority of your array is used for data storage and only a small portion is used for parity storage, this is a fairly efficient system. You'll get high performance and transfer rates on random access intensive applications.
RAID-3 gives you the capability to reconstruct data lost from a data drive but not the assurance of redundancy. If your parity disk and a data disk fail at the same time, you will lose all data and the ability to reconstruct it.
Segment Striping with Single
Drive Parity
RAID-4 is identical to RAID-3 except that large stripes are used, allowing disk read operations to overlap. This gives you a small increase in performance when accessing large files, but no other significant benefits.
Segment Striping with
Distributed Drive Parity
RAID-5 stripes blocks of data as well as parity information across all drives in the array, ensuring that no data will be lost in the event of a single drive failure. This "rotating parity array" also avoids the write bottleneck caused by single parity drive configurations in RAID-3 and RAID-4. Under RAID-5, each drive takes turns storing parity information for a different series of stripes. This rotation eliminates the dependency on a single parity drive, like those used in RAID-3 and RAID-4.
RAID-5 has one of the highest sustainable data rates for reading available information. The majority of your drive space in the array will be used for saving data. Only a small amount of space is used to deliver this high level of fault tolerance. For example, with five disks in a RAID-5 array, 80% of your disks are usable for data and information, while 20% will be used to store parity stripes.
RAID-5 is well suited for transaction processing applications such as databases, Web servers, e-mail and Internet news servers. It provides a solid foundation for disk level fault tolerance. Under RAID-5, recovery is as simple as replacing the failing disk component. When "hot swap" disk components are used, your server can stay up while you change out the failed component with no interruption of service.
Copyright Bloom MicroTech Corporation
1996-98
Send comments to webmaster@bmtnet.com
Main Page | Products & Services | Corp & Govt Section | Networking
Last Updated, Tuesday, December 01, 1998