IBM Bolsters HyperSwap to Protect IBM i Against Downtime
May 2, 2016 Alex Woodie
A new release of the HyperSwap feature in IBM‘s PowerHA Enterprise Edition will give IBM i shops better protection against IBM i downtime–in particular during planned downtime events, like operating system upgrades. IBM also released a statement of direction to bring more disaster recovery (DR) capabilities to HyperSwap, including support for a remotely located third copy of IBM i data.
HyperSwap is a data resiliency feature introduced two years ago with the launch of IBM i version 7.2. The mainframe-class technology relies on the peer-to-peer remote copy (PPRC) synchronous data replication protocol (i.e. Metro Mirror) to replicate DB2 for i data between two DS8000-series arrays sitting in the same data center or campus. In the event of an unexpected failure of the primary array, HyperSwap automatically switches the production IBM i LPAR from the first DS8000 array to the copy sitting on the second array.
Up until now, HyperSwap has only been usable with entire IBM i systems. Customers could rely on HyperSwap to replicate the data for an entire IBM i system from one DS8000 array to another. It provided protection at a specific part of the stack–the storage component–which some IBM i shops found useful. But its usefulness was limited.
With the refresh of PowerHA SystemMirror for i 7.2 Enterprise Edition, which IBM announced with the introduction of IBM i 7.3 on April 12, Big Blue is fleshing out HyperSwap with capabilities that should make it more relevant to a wider audience.
Specifically, IBM is now giving HyperSwap support for data stored in independent auxiliary storage pools (iASPs). It’s also supporting LUN-level switching, enabling PowerHA to switch IBM i applications from accessing data from either nodes in a cluster of two DS8000 arrays. This feature has been available as a tech preview since late last year, and now it’s generally available.
Brian Norland, who works in IBM i PowerHA external storage development, explained the significance of the new features to IT Jungle.
“HyperSwap is a technology that provides you with protection against. . . some sort of outage with your external storage device,” Norland says. “While that’s great, it’s also nice to have your protection for your IBM i, so if you want to do any PTFs or upgrades on there, you can do it to one system while still being available. So what we now allow you to do is to combine HyperSwap with our LUN level switching, where we’re able to take an iASP from one IBM i and switch it to another one, while you’re doing your upgrades to the one you switched away from.”
If all goes as planned, IBM i applications–and the users that rely on them–will scarcely notice anything different during these HyperSwap events. If one DS8000 storage array (and all the data it contains) needs to be taken offline for maintenance, the same IBM i machine can keep serving data, even if the physical storage array that’s being accessed is somewhere else in the region.
Failovers with HyperSwap are said to take just several minutes, which is significantly less than can be expected with other technologies that require switching application servers, Norland says.
“That’s the advantage of this–when you have a storage outage you can keep running as if nothing happened,” he says. “A HyperSwap failure where we do a switch that’s unplanned would be similar to our MultiPath failure technology, in terms of what you can expect for what kind of disruption you might see. There might be a few hangs here and there, but you don’t have that big outage.”
That brings HyperSwap one step closer to becoming useful for enabling high availability in the event of a disaster (planned outages are much more frequent than unplanned ones, Norland says, which is why IBM focuses so much on enabling operations to continue during planned downtime).
But because HyperSwap uses PPRC, which is a synchronous replication, protocol, the two DS8000s must be within 25 miles of each other, or else the replication will go berserk, creating out-of-synch database updates, gobbling up CPWs and, we assume, all the doughnuts in the break-room.
That brings us to another aspect of IBM’s April 12 HyperSwap announcement: The all-important “statement of direction.”
In the announcement letter, IBM said: “IBM plans to introduce the capability to add a third system connected to the PowerHA for i HyperSwap pair via either a Metro Mirror or Global Mirror link.”
Norland describes the importance of this. “In the future we’re also looking to add on the ability to put GlobalMirror on top of HyperSwap so you can have a DR site further away,” he says. “The idea is you have DR site further away to kind of give you that complete solution.”
Norland would not commit to a timeline for delivery of this solution, which would enable a third copy of your live data to be stored just about anywhere on the globe, network bandwidth permitting. (Suffice it to say, if IBM’s development and release cadence holds up, the company would likely deliver a tech preview of this within 12 months, although that is purely speculation.)
But wait: there’s yet a third piece of HyperSwap news! Buried in the hardware-related enhancements to IBM i 7.3, IBM announced that HyperSwap is now supported on the SAN Volume Controller (SVC) and Storwize storage arrays, which are considerably less expensive than the DS8000-class machines.
However, the iASP swapping capability of HyperSwap will not be available on the SVC and Storwize arrays; it’s only supported on DS8000 arrays, Norland says. Customers can, however, use HyperSwap with SVC and Storwize to protect their IBM i systems as a whole (in terms of the database, of course, not the applications).
If all this seems confusing, that’s because it is. Virtualization at the system and storage levels have given IBM customers more capabilities, but also increased complexity. While the IBM i architecture was built with a notion of outward simplicity, the movement toward virtualization is damaging that simplicity.
Amid all this, above all, IBM wants to give customers options, and HyperSwap is nothing if not that.
“It’s one more piece of the puzzle. As you’re looking at all our technologies with PowerHA, what outage protections are you covered against for each one of these technologies?” Norland says. “Our iASP-based replication and switching technologies are protecting you against a server or an IBM i outrage, and then HyperSwap is protecting you against a storage outage. So what this is doing is really combining the best of both worlds.”