IBM’s New HyperSwap a Head Scratcher for Some
June 17, 2014 Alex Woodie
One of the new features that IBM added to its PowerHA suite with the launch of IBM i 7.2 is HyperSwap, which allows a production IBM i LPAR to be moved from one DS8700 SAN array to another and restarted almost instantly. While HyperSwap protects against the failure of a SAN in a single data center, its inability to work in a geographically dispersed manner had led some in the IBM i HA community to question its usefulness in a true disaster.
HyperSwap is a data resiliency feature that IBM is making available to IBM i shops as part of its new PowerHA Express Edition offering. The technology, which IBM originally developed years ago for the System z mainframe, uses the peer-to-peer remote copy (PPRC) synchronous data replication protocol, the same protocol used in the Metro Mirror feature that’s available with PowerHA Enterprise Edition.
The PPRC protocol replicates data very quickly between two DS8700 storage networks, which are the only storage devices to support HyperSwap in IBM i environments at this point. Because PPRC works at a subsystem level, HyperSwap (and Metro Mirror) are transparent to the IBM i host attached to the IBM storage device. In the event of an unexpected failure of the primary DS8700, HyperSwap will automatically switch the production IBM i LPAR to the secondary DS8700 device. The role swap happens very quickly; maximum time is two to three minutes for both planned and unplanned outages, according to IBM.
While the PPRC protocol gives HyperSwap (and Metro Mirror) very fast replication and role swap times, its synchronous nature limits it to working within a single geographic area. IBM recommends keeping both ends of a Metro Mirror solution within a 25-mile radius to prevent latency from increasing to intolerable levels, and we’re betting the same limitation applies to HyperSwap.
Beyond the 25-mile line, IBM recommends PowerHA customers use an asynchronous replication protocol, such as Global Mirror (available in PowerHA Enterprise Edition) or geographic mirroring (available in PowerHA Standard Edition). Geo mirroring is an IP-based protocol that’s used to mirror the contents of independent auxiliary storage pools (IASPs) from one IBM i server to another; it doesn’t require an IBM storage array, such as Metro Mirror, Global Mirror, or HyperSwap, and can be used to provide high availability between two IBM i servers using internal DASD.
If all this sounds a bit confusing, you’re not alone. Seasoned pros in the IBM i high availability space had trouble fitting it all together even before HyperSwap hit the market. Now that HyperSwap is here, the level of confusion has gone up even more.
As the vice president of product strategy at Vision Solutions, part of Doug Piper’s job is keeping up on all the technology. But during a recent partner conference, he got a bunch of empty looks when the topic of HyperSwap came up. “They had heard about it, but were all really confused about what it was and what it did and how it relates to what we do,” Piper tells IT Jungle.
Vision Solutions competes with IBM and PowerHA. As the biggest provider of logical replication solutions that provide high availability at an application level, it has a lot to lose if people ditch its MIMIX HA and iTera HA products and switch to IBM-supplied solutions. And so you don’t necessarily expect Vision to say nice things about its competitor’s offerings.
But at the end of the day, thousands of IBM i shops around the world rely on Vision and its global partner network to build and implement high availability solutions that protect their businesses. Many of these customers have chosen to incorporate parts of the PowerHA strategy, so it’s important for Vision and its partners to know about PowerHA and speak knowledgeably about how it fits into an overall HA framework.
And according to Piper, HyperSwap doesn’t really fit in–at least if it’s HA and DR you’re after.
“We define it as a class of non-IASP, non-cluster replication solutions. It’s just replicating agnostically at the storage level,” Piper says. “It’s a single-system, mainframe-based storage technology. It’s something you should think of as akin to, and in in addition to RAID. It has a very fast, almost unnoticeable swap time to a second DS8700 using synchronous replication. It’s really there to provide support for planned and unplanned storage outages and events. That’s pretty much it.”
For people who feel they need an extra layer of storage redundancy, then it’s a great thing, Piper says. “Even some of the IBM people we’ve talked to in the field, some of our old friends, are kind of scratching their heads,” he says. “If you’ve already got all of the existing redundancy and disk technology, why is it critical to have yet another layer of redundancy on the storage? How much more value is that. There may be people who have good answers.
“And I’m sure people will find value with it. I just wouldn’t want people to be confused in thinking that, ‘Hey, I’m going to go do this HyperSwap technology, and now I don’t need a DR or an HA solution,’ because that couldn’t be further from the truth.”
Piper takes issue with IBM’s marketing on the product up to this point, and he has a point. For starters, HyperSwap is sold via the PowerHA Express Edition packaging. Express editions usually are the lowest rungs in the product totem poles, but the fact that HyperSwap requires DS8700s means only the biggest Power Systems shops in the world will be able to use it. (Of course, that could change, if IBM decides to make HyperSwap available to smaller DS devices and its popular Storewize solutions.)
IBM positions HyperSwap products as the “foundation for a class of HA and DR solutions,” which is slightly dubious and provides IBM an excuse for why it isn’t providing HA or DR now. One could make the argument that it is providing high availability protection against a DS8700 failure. It’s possible that IBM’s biggest, most expensive storage arrays do fail from time to time, but it wouldn’t be in the business long if they suffered total failures on a regular basis.
IBM’s claim that HyperSwap provides DR protection is on thinner ice because of that 25-mile limit. “HyperSwap doesn’t help you in the event of a site failure. So in terms of super protection, it provides zero in added protection if you have a site-type outage,” Piper says. “It could very well be a foundation for something they have in mind for the future. But today it’s basically providing protection for planned and unplanned storage events on a local-campus basis.”
At the end of the day, Vision wouldn’t be in business long if it touted its competitor’s products. But this flogging of HyperSwap, as it’s currently defined and positioned by IBM, seems well deserved.
IBM Debuts ‘HyperSwap’ With PowerHA Express Edition