The drives were first announced at the same time HP was introducing the 5.5 release, but the new mass storage systems need PowerPatch 1 of the 5.5 MPE/iX release (HP's number C.55.01). Without the PowerPatch, the new drives simply won't be available to the HP 3000s.
HP's arrays are manufactured by the Claarion division of Data General, but carry the HP label in an OEM arrangement. Service and support is provided through HP's regular support channels for the devices, which are priced at under $1 per Mb of storage. That translates to around $21,000 for 12 Gb of storage in the Model 10 device.
The new drives are replacing and making HP's fiber-optic FiberLink (FL) disk arrays obsolete. The old devices provide just 25 percent of the transfer bandwidth offered by the Models 10 and 20. HP product manager Vicky Symonds said that customer requests for an array that supports the new Fast/Wide SCSI interface prompted HP to add the drives to the 3000's lineup.
"The basic technology that sits on top of the drives is faster," Symonds said of the new arrays' 20Mb/second interface. "That doesn't necessarily relate to increased performance in the customer's environment."
The HP-FL arrays also worked with a different Redundant Array of Independent Disks (RAID) scheme than the newer devices. HP-FL drives used RAID 3, where the parity of the drives was stored on a single disk. This could cause bottlenecks as every other drive in the array accessed it with parity information.
In contrast, the Model 10 and Model 20 units can be configured to work with RAID 5 or RAID 1 data protection levels. RAID 1, which HP recommends if the new drives are used as LDEV 1 in an HP 3000, provides the best level of performance because it mirrors disks directly. RAID 5 stores the parity on a smaller percentage of the disk, releasing more of the drive for customer data.
HP's Symonds said customers looking at RAID storage often have large storage requirements and cannot tolerate any interruption of service. That makes the devices a good choice for 24x7 shops in places like emergency dispatch and around-the-clock catalog and direct mail operations -- market segments where the HP 3000 continues to win new business.
The Model 10 and Model 20 High Availability Disk Arrays differ from the HP High Availability Storage System (HASS) in their level of mirroring. While HASS can only offer software mirroring -- which has some CPU overhead -- the newest arrays provide full hardware mirroring.
The potential downside to working with arrays in any midrange computer environment is a performance drain. Setting up as many as five disk drives on a single controller can cause performance bottlenecks, Symonds said. Large batch jobs can sometimes require a disk device have better cache capability than RAID devices demonstrate. For that kind of need, HP recommends the new Symmetrix 3000 disk arrays for HP 3000s.
HP will not offer trade-in credits or a promotion based on disk trade-ins to sell the Model 10 and Model 20 systems, Symonds said. HP continues to sell the HP-FL drives to companies that have already invested in the technology, need replacements and don't want to make the switch. To ensure availability, HP made a "lifetime buy" of the older devices when the manufacturer discontinued the drives, she explained.
The primary advantage of the new devices, other than the 20Mb per second SCSI interface to the 3000's bus, is their flexibility. The Model 10 unit scales from 2Gb to 40Gb of total capacity, and as much as 32Gb of that can hold customer data. That's because in a RAID 5 configuration, some 20 percent of the drive capacity is turned over to the redundancy operations by "striping" the information across the drives.
The new disk units provide the first capability for "global hot spares," which means extra drives standing by and configured to take over storage operations if one drive in the array fails. When a drive fails, the controller immediately replaces the disk with the hot spare and regenerates the data on the spare. Managers can then take the failed disk offline without taking the system down. Later, the controller copies data from the spare to the repaired drive, releasing the hot spare for future use.