One of the hallmarks of the System z and the mainframes before it has been their rock solid backup and recovery. The arrival of the hybrid System zEnterprise (z196/zBX) combining different operating systems and platforms, both virtual and physical machines, has the potential to make the backup and recovery process a little messy.
In a presentation titled zEnterprise System 196 Overview IBM takes up the question of backup and recovery in just one slide, #23. Its recommendation: a parallel sysplex or geographically dispersed parallel sysplex (GDPS). That’s basically buying a second system and replicating between them. It is a good solution but it isn’t inexpensive.
Bill Reeder, who handles enterprise sales and strategy for Linux on z at IBM, confirmed the parallel sysplex and GDPS recommendation and went on to note that tape backup, once the mainframe backup mainstay, remained a commonly used option. High availability clusters provide yet another option for Linux on z, he noted, especially when operating stateful servers. If you are running stateless, you don’t even need high availability to resume operations.
Much also depends on what level of backup and recovery you are concerned about. IBM already offers a number of data management products that provide recovery capabilities and are not affected by the presence of the zEnterprise. As software tools these products apply to a zEnterprise configuration just as they do to loosely coupled distributed systems. If the backup concerns revolve around the backup and restoration of server hardware configurations and firmware recovery then the zEnterprise Unified Resource Manager (referred to a zManager) handles it.
JD Williams, an early adopter of the z196, runs the machine primarily for Java and WebSphere applications and image and video serving. Check out the JD Williams case study here. In terms of backup and recovery, the company takes a traditional approach using an ATL to handle regular backups to cartridge. In addition, copies of the tapes are sent off site to a 3rd party storage company. Should it need to recover the z, the cartridges are sent to its DR site (another 3rd party) to begin the recovery process.
However, the company also runs a z10 as well as a z196. This opens new backup and recovery opportunities. “We will be looking at the role of both the z196 and z10 with a view to a failover capability,” says Jeff Cattle, head of the company’s computer service. This probably will begin take shape later this year and into next year when the company reviews its third party DR contracts.
The z196/zBX backup issue recently flared up on a LinkedIn discussion. At one point David Boyes, key developer of the openSolaris on z project, now IllumOS on z, weighed in with a completely different take on the zEnterprise DR issue. His suggestion is to leverage the zEnterprise, especially the zBX, by tapping a hypervisor like VSphere 4 (or even Xen) and take advantage of mobility to move external virtual machine systems to a zBX for DR. “In one stroke, that gets you supported Linux, Solaris x86, and Windows workloads that can be Vmotioned directly into the zBX box (plus the Power workloads). Now you’ve got a very flexible DR platform.” Sure, simple as pie.
For years companies opted for the mainframe because of its extremely high resiliency and its rock solid backup and recovery. With the zEnterprise that hasn’t changed except that it does open some new DR possibilities to think about.