3000 High Availability takes extra steps

HP extends Fiber Channel links, adds low-cost disk storage, XP support

Hewlett-Packard used its Webcast capability to announce new High Availability (HA) products and features for e3000 customers, bundling improved High Availability Failover (HAFO) software in MPE/iX and linking switches and hubs with Fiber Channel.

The introductions unveiled in the one-hour Webcast are designed to give the e3000 better protection for its data paths and further protect its reputation for reliability. HP’s announcements also covered a new DS2100 disk storage unit and an ahead-of-schedule support introduction for the XP 48 and XP 256 disk arrays.

The new piece of hardware for the systems is an $800 SureStore DS2100, a rack-optimized, 4-slot disk system offering Ultra 2 SCSI interconnect speeds. The device allows sites which use the Just a Bunch of Disks (JBOD) storage strategy to reduce the amount of space required for significant amounts of storage. The DS2100 can hold more than a quarter of a terabyte of storage in 72Gb disks in a “1U” size rack space of 1.7 inches high, 18 inches wide and 15 inches deep. HP says that incremental storage can be added so up to three enclosures can be daisy-chained on a single channel, providing up to 12 slots in a 3U space.

Justin Jones, High Availability Product Marketing Manager for the e3000 division (CSY), said at the Webcast the DS2100 offers owners of 3000s who use internal disks a way to increase reliability while expanding disk capacity.

“It’s low cost, high performance, and runs on the new LVD-SCSI bus running at 80Mb per second,” Jones said. “Customers looking to add high availability to a solution, Mirrored Disk is an excellent choice for this, too.”

Lower-end customers buying new A-Class systems or using internal disks for LDEV1 on smaller HP 3000s are being targeted for the DS2100, Jones said, “smaller shops with just one or two 3000s.” Adding storage capacity will be easier and less costly with the DS2100 for such customers, he explained.

While the DS2100 supports Ultra 3 and the pending Ultra 4 disk drives, HP is only offering Ultra 2 controllers for its new N- and A-class e3000s. The Ultra 2 controllers work fine with the 2100, but don’t provide all the performance available if the 3000s were able to use an Ultra 3 or Ultra 4 disk.

HP also announced immediate support for the low-cost XP 48 and high-end XP 512 SureStore disk arrays on the HP 3000. The RAID storage units are being supported on the base 7.0 release of MPE/iX, instead of making customers wait for the Express 1 release of 7.0.

Fiber Channel hubs, switches

HP also announced it’s expanding the connectivity for the Fiber Channel peripheral bus on the e3000s, although native Fiber Channel is still some time away. HP is adding switch and hub support to the system’s Fiber Channel solution.

“Customers were stuck with a one-to-one connection scheme to their storage device,” Jones said, “going from the SCSI bus into a Fiber Channel connection on XP arrays. For every port coming out of your server, you’d have to use up one port on your storage array.” For example, on the XP512 there are 32 ports, “so many customers were running out of ports, and having to buy extra XP arrays. This will save customers the extra cost of having to buy another array, by allowing them to buy a switch or a hub instead.”

In the new configuration, a single connection can go from the array to the hub, and that hub can be used for multiple ports to the 3000 or multiple 3000s linked to the arrays. HP recommends a 4-1 connection topology.

Return of HAFO, Cluster/iX

HP is reintroducing High Availability Failover software for the e3000, after the first releases proved to be less than optimized for customers’ use. HAFO protects from a failure in the IO path through a cable or disk card, eliminating a single point of failure, by letting system managers configure an alternate path to the storage.

HP demonstrated HAFO for the first time several years ago, “But after we took a look at how we were implementing it, we decided it wasn’t ready for prime time,” said CSY storage engineer Walt McCullough. “The first implementation required a dedicated, unused channel or bus to act as a spare tire. That was an inefficient use of the equipment.”

McCullough said HP developed a new HAFO “so both channels can be used as if they are a spare tire that’s already on your drive shaft and ready to go.” Customers can now configure other disks to the alternate path at the same time it’s available as a spare. HP said this provides “the added benefit of bandwidth of that second path, so it’s good for performance.”

When a primary path fails, the HP 3000 notices IOs have stopped and will mount the volumes on the secondary path. Combined with the array protection in XP devices, HAFO/iX protects from the disk all the way to the e3000. HAFO/iX is being included in the MPE/iX 6.5 Express 2 and the later 7.0 Express 1 for HP 3000s at no extra charge. It works with the XP 48, XP 256 and XP 512 arrays.

HP also announced a new version of Cluster/iX, the extra-charge product designed to protect from loss of access to storage and applications when a primary system fails. The new version is independent of any disk technology and can use any dual-ported array. Its predecessor was confined to using only certain disk drives from HP. Problems surrounding boot up processes have been eliminated, and the new Cluster/iX now handles more than two computers.

When a primary system fails, Cluster/iX lets an operator grab control of the console of the cluster’s volumeset and take it away from the primary. Applications on the secondary system are restarted and users are logged on the secondary system. Primary and secondary systems must have their own system volumeset.

HP is also allowing for disaster recovery where the HP 3000s are in different locations. The new Fiber Channel capabilities of the XP 512 and XP 48 arrays mean those devices can be 500 meters distant from HP 3000s. Managers can also use continuous access for the XP arrays, which gives up to 15 km distance between arrays. The primary system can be connected the first XP 512, and the secondary system connected to the second XP 512 at another remote site.

The revised Cluster/iX is offered on MPE/iX 6.5 Express 2, and will be available for the new e3000s with the Express 1 release of 7.0.

HA strategies

Several industry experts from the e3000 community joined HP in the Webcast to outline strategies for high availability on the systems. Below the capability of Cluster/iX and the redundancy of disk arrays lies Mirrored Disk/iX, a software solution which mirrors one disk to another. HP has been considering work to give the software the ability to mirror the boot-up volume for e3000s.

In a follow-up question and answer session, Jones said in the second half of 2001 HP will add support for the Virtual Array family of devices, the va7100 and va7400 arrays to update HP’s SureStore 12H AutoRAID units. The Virtual Arrays offer Fiber Channel connectivity not available in the 12H devices.

HP said the most cost-effective way to implement Mirrored Disk is to use the new DS2100. Jones said for about $7,500 customers can get two IO cards, two cables, two DS2100s and two 18Gb disks. Buying an array protects all storage volumes, while using HAFO to protect all points of failure in the data path.

Webcast host George Stachnik said he’s seen customers “lull themselves into a false sense of security” and believe there’s no reason to do regular backups. Denys Beauchemin of HICOMP outlined the four types of system backup and talked about cross-platform backup solutions which include the HP e3000 as either a controlling system or a backed up system.

Single-system backup, networked backup servers, LAN-free backup and Storage Area Networks make up the four types of backup, Beauchemin said. The first is the most common to the 3000 community, but “As you get more systems, you have to bring in more tape devices, so that can get very expensive,” he said.

In contrast, a networked backup solution gives sites a way to move up to high-end storage devices such as libraries, sharing such devices. A LAN-free environment shares across differing platforms, but is unbound by the speed of the LAN. “The downside [of LAN-free] is that the library’s tape drives are dedicated to the systems,” Beauchemin said, “so they cannot be used by another server.”

Storage Area Networks share devices within libraries, to maximize the use of high-end hardware. The HP e3000 can “absolutely play in the SAN environment,” Beauchemin said. SANs can reduce the number of tape drives needed, but their start-up cost can be high.

HiBack Software from HICOMP connects the e3000 to SANs, as well as multiple platform environments. The e3000 systems will have to connect via a fiber bridge, the hardware component HP announced in the Webcast it has added to the 3000’s Fiber Channel capability.

VP of Technology Mark Klein of ORBiT Software also offered a view of technology new to most e3000 shops during the Webcast. Klein explained the company’s Delta backup process and Online Zero DownTime backup features of ORBiT’s Backup+ product.

With Delta, “we’re backing up only changed pages of disk,” Klein said, making backups shorter and conserving capacity. “Today we have large files on the HP 3000; the maximum dataset size is 80Gb. If you modify a gigabyte’s worth of data, [with incremental backups] you still backup 80Gb of data just to get the 1Gb of changes. You can get a tremendous time and space savings by simply backing up only the modified activity with the Delta technology.”

Zero DownTime keeps e3000 managers from having to restart applications after backups. “The applications can take 20-30 minutes to start up, and 20-30 minutes to shut down,” Klein said. “That’s a very expensive prospect to stop and start the application.”

ORBiT believes “the hardware is the limiting factor of the backup, and we’ve been able to prove that is the case through our testing on the N-Class systems. As the hardware gets faster, Backup+ will be able to scale with little effort on our part.”