7.11. RAID AD-Advanced Data Guarding

7.11. RAID AD Advanced Data Guarding

RAID advanced data guarding (ADG) delivers high fault tolerance similar to RAID 1, but keeps capacity utilization high, as with RAID 5. It protects data from multiple drive failures with an ability to withstand two simultaneous hard drive failures without data loss or downtime.

RAID ADG uses two sets of parity striped across the disks, as shown in Figure 7-12. This method results in protection for an array with as many as 56 drives and requires the capacity of only two drives to store parity information.

Figure 7-12. RAID ADG.

7.11.1 RAID ADG Performance

ADG employs two independent parity schemes. Two sets of parities must be modified for each logical write operation, which requires six physical transfers three reads and three writes.

After the existing data and both sets of parity are read from the physical drives, each set of parity is modified by operating on the existing data, the data to be written, and the old parity. The new data and both sets of modified parity data are then written to the physical drives.

Although RAID ADG provides the dual advantages of increased fault tolerance and high capacity, it does so at the cost of performance that is less than that of other RAID levels.

Performance equals that for RAID 5 when reading data, but is slower when writing because of the extra parity data.

7.11.2 When to Use RAID ADG

RAID ADG delivers high fault tolerance similar to RAID 1+0, but keeps capacity utilization high, as in RAID 5. This higher level of protection is ideal where large logical volumes are required.

Advantages of ADG include the following:

• It safely protects an array of up to 56 total drives; extensibility increases the number of parity drives to provide greater fault tolerance.

• It can tolerate two simultaneous drive failures without downtime or data loss.

• It has greater fault tolerance than RAID 4, RAID 1+0, or RAID 5.

• It offers lower implementation cost than RAID 1 or RAID 1+0.

• It supports online spare drives.

• It supports online RAID level migration from RAID 1, RAID 1+0, or RAID 5.

• It is ideal for applications requiring large logical drives with many physical drives.

RAID ADG is best implemented under the following criteria:

• When protecting mission-critical data

• When greater capacity utilization is needed

• For applications requiring larger storage volumes

• When better capacity than RAID 1 or RAID 1+0 is needed

• When higher fault tolerance than RAID 1, RAID 1+0, or RAID 5 is necessary

7.11.3 HP ADG Support

For ADG support on a ProLiant server, use the HP Smart Array 5304/128 controller with at least four disk drives, or use the HP Smart Array 5302 controller with a minimum of 64MB cache and the ADG upgrade kit.

HP recommends RAID ADG or RAID 1+0 for more than 8 disks. RAID 5 is recommended for up to 8 disks (maximum of 14).

7.11.4 RAID ADG Fault-Tolerance Comparison

Figure 7-13 shows the relative probability of logical drive failure for different RAID settings and different logical drive sizes, assuming no online spares are present.

Figure 7-13. Logical drive failure probability for different RAID configura-tions.

With growing numbers of individual disk drives needed in a single logical volume on a single controller, RAID ADG provides higher data reliability and has a superior fault tolerance when compared to RAID 5 or RAID 1.

With RAID 0, the logical drive fails if one physical drive fails.

With RAID 5, the logical drive fails if two physical drives fail.

With RAID 1 or 1+0, the maximum number of hard drives that can fail without failure of the logical dive is n/2. However, a RAID 1+0 logical drive fails if only two hard drives fail, if they are mirrored to each other.

With RAID ADG, three hard drives must fail before data loss is incurred.

For RAID 5, HP recommends that no more than 14 (8 is optimal) physical drives be used per logical drive. However, logical drive failure is much less likely with RAID ADG, and HP supports the use of up to 56 physical drives per drive array when running this fault-tolerance method.