The most important disk drive facts are...
GigaBytes per Cubic Inch As drive sizes decrease, the amount of data that can be stored in a cubic volume (GigaBytes per Cubic Inch) increases exponentially as illustrated below. GigaBytes per Cubic Inch as a Function of Disk Diameter A disk database system implemented using 2 inch drives will consume much more space than the same database implemented using 1 inch drives. Watts per GigaByte As drives are made smaller, the number of watts per unit of data (Watts per GigaByte) consumed by the drives decreases exponentially as illustrated below. Watts per GigaByte as a Function of Disk Diameter Most of the power consumed by an HDA is consumed by the motor in spinning the disks. Rotating disks are like a fan, there is a lot of aerodynamic drag, and the inertia of a disk varies as the fourth power of radius. In order to maintain the same degree of controllability, the power consumed by the motor also must increase as the fourth power of radius! Think what that means for disks of diameters 2 inch and 4 inch (14 = 1 and 24 = 16! ). In other words, a 4 inch HDA will probably consume more than 10x as much power as a 2 inch HDA and a 2 inch HDA will probably consume more than 10x as much power as a 1 inch HDA. This fact is extraordinarily significant! Consider a disk database system implemented using 2 inch drives which consumes 1,000 watts of power. The same database implemented with 1 inch drives would probably consume less than 100 watts. Two Main Classes of RAID All RAID systems can be divided into two main classes - those where the drives in a drive group are written and read independently (RAID-4, RAID-5 and RAID-6) and those where all of the drives in a group are written and read simultaneously (RAID-1, RAID-2, RAID-3 and ECC Tek’s 2D-RS RAID). Because of the write overhead penalty in RAID-5 type RAID, it simply is not practical to implement RAID-5 type systems if failures occur frequently. The write overhead inefficiency is too great and increases as the amount of redundancy is increased. The performance of large RAID-5 systems rapidly deteriorates catastrophically if failures occur frequently or if a large amount of redundancy is used. Future Disk Databases Future disk databases will use very small drives (as small as practical) because smaller drives consume much less space and much less power than larger ones. A sophisticated error correction system must be implemented because disks will fail fairly frequently. In order for the database to be reliable and always be available with high performance, frequent failures must be tolerated without any significant degradation in performance. Only the RAID-2 class of RAID (which includes ECC Tek’s 2D-RS RAID) can correct for frequent failures with no loss of performance, and the 2D-RS RAID scheme is the only RAID scheme needed because it can be configured to emulate every other RAID scheme. |