Dell Enterprise RAID and Physical Drive: Replacement FAQ - Can different drives be used in a RAID?

Dell replaces hard drives under warranty with drives of similar or better specifications. Replacement drives may be larger or faster and may also be from a different manufacturer. Dell's replacement drives are certified to be compatible with Dell hardware.

If the failed disk is no longer under warranty, you must get a new drive.
For MD Series, you find the list of supported Physical Drives in the Support Matrix of the associated storage enclosure. 

• If using a JBOD connected to an MD RBOD, use the MD3 RBOD Support Matrix

Warning: Only the physical disks listed in the physical disk table with Dell Part Numbers listed in the Support Matrix is supported and shown as "Certified Drive."

 
Explanations:
It is possible to buy from an external provider a disk listed in the supported table, for example "ST9500430SS" without the Dell certification, but Non-Certified Drive does not work in an MD array.
To be sure the Hard Drive is Dell certified, the drive must have a Dell Part Number, it means that the disk is manufactured to Dell’s standard, runs a Dell version of firmware, and is fully compatible with MD arrays.
Generally there is a Dell sticker on it.

No.  It is perfectly valid to use hard drives from different manufacturers, model numbers, sizes, and rotational speed (spindle speed or RPM).

Using identical hard drives was never a requirement within a RAID array since the initial concept of RAID. However, there are several factors that are likely contributors to the perpetuated myth that hard drives must be identical.

The first is that from the perspective of initially creating a RAID array using hard drives, it does not make practical sense to use dissimilar drives. For example, if you choose to combine a 500GB drive with a 100GB drive, you can only use a maximum of 100GB of the 500GB drive in the array resulting in a net loss of 400GB of space. If you choose to use a 10k RPM drive and a 15k RPM drive it works, but you would lose the advantage of any faster drives (generally). Given the likely cost differential between the larger and or faster drives with no practical technology benefit, it makes little fiscal sense to design a RAID array in such a way, although there is no technological reason restricting such a design.

From a hard drive replacement strategy perspective, replacing a 100 GB drive purchased in 2007, with a 500 GB drive purchased in 2010 is often the best choice. 100GB hard drives are likely to have become rarer, and could actually cost more than a 500GB drive, although the 500GB drive may have superior specifications (in size and speed). Since it does no harm to use the larger (and possibly faster) drive, if this is the most cost effective option, it is the best choice.

Second, at the time when RAID was invented (in 1988) and well into the 1990s, hard drive firmware was relatively immature. It was common to experience compatibility issues when using multiple hard drives. For example, using Hard Drive A by itself works and using Hard Drive B by itself works. But when Hard Drive A and Hard Drive B are used together, connected to the same controller, often many issues were encountered. This issue was so prevalent that it becomes a Best Practice at the time to use identical drives (same manufacturer, same model number, same lot number, and so forth) when using multiple drives whether using them in a RAID array or not. Hard Drive technology has matured to the point that such concerns are no longer applicable, and have not been applicable for over a decade. 

Given the first example (fiscal reasons for using similar drives when creating an array) and the second (compatibility issues), it may be easier to see how the myth of using identical drives has propagated over the years and continues to be propagated today.    

None. It is generally enforced that hard drives must use the same interface technology (SCSI or SAS or SATA). Also, it is also generally enforced that mechanical drives (hard drives) and non-mechanical drives (SSD) drives cannot be mixed in the same array. While it is not technologically prohibited to mix SAS with SATA or SATA with SSD or SAS with SSD within an array, the potential for confusion regarding configurations and performance variables is significant with little to no practical benefits. These facts cause most (if not all) manufacturers of Enterprise RAID technologies to disallow such combinations.

Inherent in RAID technology and design is the concept that hard drives eventually fail and require replacement. Given the continual advancement of technologies, it is understood that identical hard drives may be difficult, expensive, or even impossible to acquire at the time replacement is needed. RAID technology, since its original design, never required identical drives. 

More specifically, every hard drive attached to a controller (RAID or otherwise) always operates independently of every other hard drive attached to that controller. RAID controllers logically combine physical drives into arrays, however, from a hardware communication perspective separate commands are sent from the controller to each individual drive. Even hard drives that are identical (manufacturer, model, specification) are not likely to operate at identical speeds always. 

Even when using drives of dramatically different speeds and sizes, RAID functionality continues without risk of data corruption or loss due to the differences in the drives.

RAID controllers allocate space on hard drives into arrays by the smallest drive in the array. For example, when creating a RAID 5 with three drives sized 500 GB, 200 GB, and 100 GB, each drive has a maximum of 100 GB allocated towards the array. Any excess space on the larger drives is not usable within that array, and may or may not be available for use in other arrays.

A RAID 5 array created with three 500GB drives could use 500GB from each drive. When replacing one of the drives with a larger drive (2 TB for example), 500 GB would be allocated array during the rebuild and any additional space would be unusable within the array.

RAID controllers communicate to drives individually. When completing an I/O to a RAID array that consists of multiple drives, each drive receives separate commands that are completed at the speed of the individual drive. When all drives have finished, the I/O is then considered complete.

For example, with a RAID 5 array with three drives, assume drive 0 takes 20 ms to complete an I/O, drive 1 takes 15 ms to complete an I/O, and drive 2 takes 10 ms to complete an I/O. The faster drives finish sooner, but the controller does not send additional commands until the I/O is complete to all drives. In this example, the total I/O time is 20 ms (the speed of the slowest drive).

For this reason, adding in drives that are faster, such as replacing a 10k RPM drive with a 15k RPM drive does not adversely affect the operation of the array. Replacing faster drives with slower drives may reduce the performance of the array and for this reason drives replaced under Dell's warranty are similar or faster speed. But in either case, there is no risk of corruption or data loss as a result of mixing drive speeds within an array.

Drives that are smaller cannot be rebuilt into an array. Drives that are slower may adversely affect the performance of the array. Drives that use a different interface specification (SAS, SATA, SSD) cannot be used.