PowerHA Implementations No Picnic; Help On The Way
September 28, 2016 Dan Burger
Easy as pie does not describe the implementation process for PowerHA. It can be a mind-bender, even in shops where the staffing is adequate and the skills are current. So how about something a little easier to implement and why not make it applicable to all three operating systems that run on Power? It’s coming soon, even though the official announcement has yet to be made. Easier implementation should be a big boost to PowerHA.
At least that’s what Satya Sharma, CTO and IBM Power Systems Fellow, told me when we met last week at the Edge 2016 conference in Las Vegas. Although I’ve never had anyone tell me their product is directly related to mental breakdowns, I let Sharma explain how GDR (an acronym for Geographically Dispersed Resiliency) would be a simplification of a process that probably couldn’t be made any more difficult.
The details on this are slim, we’ll know more when the official announcement is made, but there was some chatter to be heard about this at the Edge 2016 conference last week. Sharma described the essence of GDR as depending on VM restart principles rather than PowerHA’s requirement of creating a cluster member. The VM restart technique, he says, takes a little longer to recover, but it is much easier to implement.
He also noted a break from tradition.
Previously HA was ingrained in the operating systems and the similarities between PowerHA for AIX and i were minimal. Each had its own separate code base. GDR was designed at the platform level. In that regard, it is patterned after the Power hardware and PowerVM. It has identical tooling for i, AIX and Linux.
“This is a big deal. It’s the single biggest Power Systems investment made this year,” Sharma said. “Linux is partially the reason for this. To support Linux alongside IBM i, the more that things can be done in an OS-independent fashion, the easier it becomes to support all three OSes. Otherwise the investments get made three times, once for each OS.
“IBM i is riding on the Power Systems investments, which is better than when it was a unique platform and investment was tied to that unique brand. The investment in Power Systems is much larger. It’s propelling Power Systems forward. The sum total of IBM i, AIX and Linux business is much larger. The investment in hardware, infrastructure and virtualization is much greater than if i was standing alone.”
PowerHA is made to work with storage area networks (SANs), which have come on strong in the past five years. As usual, the large IBM i shops led the way, but smaller shops are moving to external storage as well.
Larry Bolhuis, an IBM i system admin subject matter expert, says shops running P05 hardware are starting to build SANs and that many shops running P10 boxes are already moving to storage area networks. At the enterprise level shops, he says, most shops have SANs or in the process of moving storage to a SAN.
For companies that have not already invested in software-based HA, PowerHA is a good option, according to Bolhuis. Compared to software-based HA that uses journaling and is sometimes stymied by objects that won’t be journaled, PowerHA doesn’t take into account what kind of object is being replicated.
A SAN sets up an environment to be virtualized, which means that workloads can be moved around easier and hardware can be used for backup and replication.
Industry experts like Tom Huntington at HelpSystems predict SANs will be in the clear majority five years from now in IBM i shops. “I see trends developing is storage area networks,” Huntington says.
From Sharma’s perspective, storage area networks with Virtual I/O servers are the single biggest change in storage during the past three to five years. He says large customers see centralized storage as an advantage and recent releases of IBM i have made it possible.
“Virtual I/O server, PowerVM, has an advantage because it can support a vast array of storage controllers, from a variety of vendors–IBM storage, Hitachi, EMC, and others. So IBM i has access to all these storage controllers through Virtual I/O servers,” Sharma says. “All the virtualization management tasks are easier when you virtualize I/O.”
Virtual I/O also allows the advanced management capabilities of live partition mobility, a feature that allows a running logical partition to be moved between physical machines.
Although PowerHA stands on its own and is not dependent on storage area networks and virtual I/O, it’s common to see it in these environments. Most of the credit for that comes from IBM being the guiding hand in a majority of SAN implementations and its services people favoring PowerHA, unless another HA solution is already in place.
Sharma gauges the market penetration of PowerHA by comparing it to the total number of LPARS on Power Systems. His claim is that both versions of PowerHA–one for i and one for AIX–are covering approximately 30 percent of their respective LPARs.
Getting back to the GDR product, Sharma says it will go a long way toward covering the 70 percent of LPARs that PowerHA is not covering.
“If you classify your workloads as platinum, gold, silver, copper, and bronze. The platinum workloads are the ones that are typically covered with PowerHA and they will continue to use PowerHA. The rest of the workloads are generally uncovered by either DR or HA. Those are the workloads expected to be covered with GDR for Power.”
When GDR is introduced next month, only AIX and Linux versions will be available. Support for IBM i is expected to come in Q2 2017. Although the code is the same across the OSes, Sharma says testing and QA still needs to be completed on the i version of the product.