RAID stands for Redundant Array of Inexpensive Disks. RAID uses multiple disks to increase performance and/or reliability. Many motherboards today support RAID in the forms of RAID0, RAID1 and RAID0+1. There exist other modes including RAID3 and RAID5.

Advantages of RAID include - improved performance and/or reliability

Disadvantages of RAID include - cost- multiple disks are treated as one array (2 60 gig drives for example will be treated as 1 60 gig array in a RAID configuration), reliability in RAID0 mode, efficiency- striped mode does not equate to a 100% performance boost. There is a certain overhead to RAID.

RAID0

In a RAID 0 configuration, multiple disks are used to stripe data. Striping means spreading out the data over multiple drives. Since we are writing to multiple disks we are not bound by the performance limitation of a single disk resulting in greater performance since we can a) write to multiple disks simultaneously and b) read from multiple disks simultaneously.



The problem with RAID0 is reliability. If one disk fails, the entire array fails. This does not sound too bad until you consider the fact that the probability of failure doubles when compared to a single disk. How does this work? Take a look at the probability tree below. Manufacturers claim roughly a 1% failure rate (I call shenanigans on that number but we'll believe them for now). If we just look at Disk1 everything is pretty good. 1% failure rate and that's it. But since in RAID, the probability of failure is dependant on both disks, we have to add another tree. The probability of both disks failing is low, at 0.01% which makes sense but there is the probability of either disk failing. Since there are 3 scenarios which would result in data loss, we sum up the probability of failure which would be 0.01% + 0.99% + 0.99% which gives us 1.89%. That's a fair size jump from 1% but in most cases we are ok. This rule holds more or less consistent regardless of failure rate. A 3% chance of disk failure would result in roughly a 6% chance of lost data in a RAID0 array. With more disks, the probability of error increases, we just add another layer to the tree, with 3 disks at 1%, we're up to a 3% failure rate with 4 disks 4%.

RAID1

In a RAID1 configuration, multiple disks are used to mirror data. Mirroring means that the contents of the disks on the array are identical over the RAID volume. The contents of Disk1 should be the same as the contents of Disk2. The disadvantage of RAID1 is the same data must be written to multiple disks so we do not get any performance increase when writing. When reading, we have the benefit of multiple disk sources that can be read from so that should be faster than a non RAID array. Reliability is better in a RAID1 array- back to the handy probability tree above, the failure rate would be 0.01% since the only way we will lose data is when both disks fail simultaneously. In any other configuration our data will still be okay because the other disk is still functional.

RAID0+1

RAID0+1 implies both striping and mirroring and requires a minimum of 4 disks whereas RAID0 and RAID1 can be done with 2 disks. 2 drives are used for mirroring and 2 for striping resulting in both reliability and performance. The biggest disadvantage of course is the overhead of having to buy 4 disks.

Is this answer useful? Yes   No So far 249 of 357 person(s) think this answer is useful.