To Optimize Your Power11 Investment, Trust Your Solutions
To Optimize Your Power11 Investment, Trust Your Solutions
A successful upgrade is about more than just the server
By Andrew Wig
The answer isn’t to max out your data center, says Randy Watson, a capacity and performance analysis expert at Fortra. “I don't see the point of spending lots of money on a server that's clearly oversized for you, when you can do the same on a smaller server,” Watson says. “And if you're afraid of having a technical problem, you just need to trust your solutions.”
Build for Failure
It’s not uncommon to see IBM Power shops acquiring far more server capacity than they need, the capacity planning experts at Fortra say. For instance, Alex Lazzaro, senior technical consultant at Fortra, has seen shops equipped with a capacity of 48 cores but only using 10. The rationale for such overbuying, he explains, is redundancy—and the reliability that confers.
While it’s true that the enterprise-class IBM Power servers are more reliable due to redundant parts, it’s typically not worth buying excess capacity just to attain that reliability, according to Watson. “It's the executives who want to feel warm and fuzzy that this machine's never going to go down,” he says.
Those decision makers should take a cue from the hyperscalers in the cloud, where they “build for failure,” Lazzaro says. Instead of aiming to use servers that allow concurrent maintenance, hyperscalers rely on redundancy at all levels (hardware, network, power supply) in case of failure.
On Power servers, some components don’t fail often enough to be concerned with redundancy, Watson notes, pointing to the system clock as one example. There are two system clocks in the highest-level Power servers, and “I can count on one hand, in the 35 years I've been doing this, that I've seen a system clock fail,” he says.
Or, take the service processors—enterprise-class Power servers have two of those, too. They can fail, Watson says, but they are hot-swappable, easing concerns about production interruptions.
- Led by Industry Experts
- Live or On-Demand Access
- Subscribe for Exclusive Invitations
Put in the Work
Too often, Watson has noticed, HA/DR is treated like a checkbox. He says backup and recovery takes more thought and care than that.
Todd Brooks, CEO of FalconStor, seconds that observation. “I think that is a universal situation,” Brooks says. “ … So many times, we just call it backup. That's only half the equation. There's the recovery part, too.”
A Power shop that doesn’t routinely test their recovery capabilities may have decades-old data on tape, and no idea if they will be able to recover it. “They may not have gone back and tested those tapes to make sure you can still read them. That's a super laborious process to do that,” Brooks says.
The smart customers, Watson says, will test their HA/DR’s effectiveness by conducting regularly scheduled roll swaps between their main data center and their DR site.
The 3-2-1 Backup Rule
Smooth Migrations
Of course, data transfer is also a critical part of the upgrade process, so make sure the network speed in your data center is up to par, Lazzaro advises. Underperforming speeds may mean disruption during migration. “If you don't have a good speed in your data center, Live Partition Mobility is out of the question, for instance,” Lazzaro says.
Because all that data takes time and money to move, upgrading your server might also be a good opportunity to clean up your disc, Watson notes. “Those transfers, they are not cheap,” he says. “Sometimes I see customers that have, I don’t know, maybe even movies on the IFS that somebody uploaded in 2003.”
Your Migration Toolbox
- Migration Preparation Tools
- Tools to Ease the Data Flow
Consider the Unsung Features
Zero Planned Downtime
Power servers are able to facilitate migrations while minimizing or eliminating disruption—a capability IBM calls zero planned downtime. “The Power platform has been capable of zero planned downtime for a long time, but the onus has always been on you/us the admins to be able to get all the things needed aligned to perform,” notes Shawn Bodily, a senior IT consultant and services manager at Clear Technologies.
With new operations tools in place, such as IBM Concert and IBM i Migrate While Active, that task has gotten easier. “The concept of zero planned downtime is not new, but the ease of how to achieve it is much greater now,” Bodily says.
Resource Groups
One of the most significant yet least understood features in Power11 is the new Resource Groups, Watson observes. Resource Groups provide a physical barrier between LPARS, preventing the “noisy neighbor” effect, in which a resource-heavy workload in one LPAR encroaches on a workload in another LPAR, despite the logical separation.
“That's a feature nobody really knows about or understands,” Watson says. “In fact, I don't think I've heard anybody use it yet.”
PEP 2
Once reserved for the larger Power servers but now more widely available, Power Enterprise Pools 2 (PEP 2) is a resource-sharing capability that allows organizations to pool compute resources across multiple physical machines to optimize licensing costs and simplify management.
“PEP 2 use is still not as widespread as I would have imagined, even more now that it’s available on the smaller servers,” Alex Lazzaro says.
PEP 2 helps with licensing by shifting the management of software and hardware resources from individual serial numbers to an enterprise-wide pool. The Fortra experts blame the underutilization on a lack of understanding, especially as it relates to licensing. “Licensing for IBM i is an obscure art,” Lazzaro says.
Adds Watson: “A lot of business partners don't understand it. A lot of IBMers don't understand it. So they don't promote it or sell it,” Watson explains.
Use Your Resources Wisely
For all the talk about the technology, the ultimate test of your upgrade’s success will be the financial results. That answer can arrive fast, Watson says: “It often doesn’t take long to recoup the investment and show gains, compared to doing nothing.”
