PowerHA On IBM i Lags Behind AIX
November 11, 2013 Dan Burger
Hardware-based high availability has been in the IBM playbook for several years, and Big Blue has made it one of its Power Systems priorities. Steve Finnes, IBM’s worldwide product offering manager for PowerHA on AIX and IBM i, isn’t one to play favorites when it comes to one platform over another. But when it comes to PowerHA, even Finnes will admit the product’s popularity on the AIX side far surpasses the embrace by the IBM i community.
In a conversation with IT Jungle at the Enterprise2012 conference, Finnes estimated the AIX users of PowerHA totaled “probably over 10,000 clients.” And he described PowerHA on AIX as “pervasive across all industries.” Typically these are AIX shops moving to storage area networks (SANs), he says.
He went on to say the percentage of IBM i shops is “relatively small by comparison–less than 10 percent of AIX–but growing by double digits.”
Describing his view of the IBM i community, Finnes said, “there are companies that have used software replication for years. There are companies moving from that solution to the clustering stuff. It’s a trend. It is happening. I’m observing it.”
He also went on to say “more than 50 percent of the PowerHA installations are conversions from software-based HA.”
These aren’t exact numbers. IBM, and therefore Steve Finnes, doesn’t provide statistics on product shipments or customer installations. These numbers are anecdotal. But after doing the math on this, I get the idea that fewer than 1,000 IBM i shops have implemented PowerHA. And, for the record, Finnes estimated that number to be more than 2,000 Power HA for IBM i deployments when talking with IT Jungle in February. Like I said, these aren’t exact numbers.
Regardless of the number of installs, PowerHA marketing has a full head of steam. The fact that it is better accepted on the AIX side has a lot to do with external storage being standard operating procedure there. Meanwhile, internal storage has been SOP on the i side. There is, however, reason to believe that external storage in IBM i shops may be gaining interest. And there are several factors that could be influential.
The first is the economies realized by multi-platform shared storage (clusters) in SAN environments. The second is the increasing availability of external storage options for IBM i small to midsize shops. And the third is the pressure created by the ever-diminishing window for completing backups and the lure of flash drives and solid state disk.
Clustered storage distributes workloads among storage servers and is implemented to realize a gain in performance and capacity. It creates a pool of disk in a storage area network. PowerHA works in this environment when the database is placed in the cluster. As Finnes points out, this allows for mirroring rather than replication. In this scenario, the data is paged out of memory and sent to both server locations. And with the database in the cluster connected to two IBM i servers, it allows users to be switched from one server to the other, in case one fails or needs to be taken out of production for maintenance. The technology allows for mirrored data to be synchronous when the distance between servers is approximately 25 miles or less or asynchronous at distances beyond 25 miles.
“With PowerHA finding out if the target box is in synch is not a concern [it usually is a concern in software-based HA] because it is synchronous to the application state,” Finnes noted. “The data gets paged out of memory and is sent to two separate arrays. It is a function of storage management. When a write command is done, the database update gets done to a pool of disks and it can be switched between systems. It can be replicated in real time.
“The app does a write. The write goes to a storage server, which writes it to the second storage server. The second storage server responds to the first, which responds to the app. The app then goes to the next write operation. That makes the data in the remote and the local are synchronous to the app.”
Clustering has gone through a long evolution. The original clustering in the IBM i world began in the 1990s. The idea of pooling disk that could attach to multiple systems and access and the automation came later.
“The LUG became the driving force,” Finnes said. “They had HA as a priority. That created a sense of urgency in getting the product completed.”
Finnes credited the concept of GeoMirroring (host-based replication) for lowering HA cost to the affordable level for smaller shops. The affordability of storage servers in the SMB is also a major benefit. Midsize companies are switching from internal disk to a Storwize V3700 and PowerHA, he said.
Replacing internal disk with storage also has the benefit of flash copy, which because of its speed allows companies to replace disk or add capacity without disruption.
“Almost all of the shops using external storage are using flash copy,” Finnes said.
The costs are primarily in the storage servers, the connectivity and bandwidth (database updates and IFS updates will drive the bandwidth), and the services to set up the system. Finnes estimated service engagements begin at about 40 hours for a simple cluster and that it could go to several hundred hours for big installations. A large customer converting from internal disk to storage would be a big installation project, he said.
Of course, when you go to storage, you are learning a new skill. You have to know stuff you did not know before.
When a company makes the decision to move to external storage, it will very likely use storage replication services, the PowerHA evangelist said.
In addition to the PowerHA in the Related Stories listing at the end of this article, you can also find an IBM Redbook titled PowerHA SystemMirror for IBM i Cookbook at this link.