Designing a Capacity Planning Playbook for IBM i

Following these key principles can help you avoid surprises as you assess your hardware needs and track system performance
By Samantha Nobles

Designing a Capacity Planning Playbook for IBM i

Following these key principles can help you avoid surprises as you assess your hardware needs and track system performance

By Samantha Nobles

Performance and capacity planning on IBM i comes down to two goals: a great, consistent user experience and efficient use of resources. Teams that track the right data, know the “normal” and make decisions against clear, strategic outcomes grounded in business realities can deliver predictable performance—and fewer surprises. Achieving that goal means proactively gathering baseline data and reading it for what matters most while designing the right approach for overall capacity goals.

To learn more about these key principles, TechChannel spoke with Eric Barsness, a principal solution architect for IBM Expert Labs and master inventor at IBM; and Maciej Wielgus, IT architect at Omnilogy, a provider of observability solutions. These experts shared their field-tested guidance about designing the right capacity and performance planning playbook for IBM i.

First Step: Capture the Right Data

If teams only look at performance after something slows down, they’re already behind. Every IT professional should maintain a minimal “starter” data set to inform IBM i capacity planning, starting with a few key “always-on” data sources. Crucial tools include IBM i’s Collection Services, Performance Data Investigator, Navigator for i and Job Watcher. “We tell clients to collect performance data 24-7, all the time,” Barsness says.

 

For benchmarking, capture samples across both business-as-usual periods and peak windows, then compare future activity to that baseline. Prioritize fundamentals. “Utilization is probably the most important,” Barsness says. Finally, validate the baseline over time. “To have a good idea of your system performance, you have to measure at many points in time, over a long period of time,” Wielgus notes.

To have a good idea of your system performance, you have to measure at many points in time, over a long period of time.
—Maciej Wielgus, IT architect, Omnilogy
SPONSORED CONTENT

Tap into Valuable Insight Across Your Systems for Free

Performance analysis and capacity planning start with historical data — data you can use to plan and configure upgrades, server consolidations, LPAR configurations, and competitive migrations with ease.

Performance Navigator is the software of choice for performance analysis and capacity planning in thousands of data centers around the world. Download the free version today and tap into valuable insight across your systems.
Tap to reveal more…
Problem Determination Analysis →
Get insight into performance issues at your fingertips instead of searching for the problem.
Graphical Data →
Analyze historical data from IBM i, AIX/UNIX, Linux, Solaris, and HP-UX.
Capacity Planning →
Performance Navigator is the industry standard for system sizing, used by organizations around the world.
No-Cost Insights →
Get disk space utilization graphs, CPU utilization graphs, and daily health check reports for free.

Read the Right Signals

To know what to do with this key data, teams need to know what “normal” looks like. Barsness advises: “It’s like looking at an MRI—everybody’s brain’s a little different. Your own differences versus your baseline are what matter most.” Context matters; key signals will vary by workload, seasonality and business goals.

 

Also, don’t overreact to headline numbers or let the wrong goals pull the team down a rabbit hole. For example, 100% CPU usage isn’t automatically bad. “Even if a partition hits a hundred percent utilization at times, that doesn’t necessarily mean there’s a problem,” especially when system priorities are set so critical work gets what it needs, Barsness says. Page-faulting is another area many teams focus on, but it rarely is completely eliminated—Barsness advises teams to evaluate the cost/benefit of the investment required to get to zero.

Turning Quality Data Into Wise Decisions
You’ve gathered the relevant data; now do this with it:
 Classify workloads
 Look at historical summaries in Navigator for i to project CPU utilization and storage needs
 Set an operational threshold based on predictable utilization peaks
 Invest in new hardware based on actual requirements, not perceived needs

Put the Data to Work

Now, with the right data and the right signals, it’s time to apply them to planning and capacity. Start by classifying workloads, then use Navigator for i’s Graph History feature to project peak CPU utilization and storage growth.

 

Preemptively track predictable utilization peaks to set an operational threshold. Translate forecasts into sensible sizing and storage/network parameters. Invest in new hardware only based on actual requirements, not perceived needs.

 

A common misstep is neglecting IBM i’s own powerful prioritization mechanism. “Designating priority is a very efficient mechanism in IBM i that allows us to stratify jobs to an unbelievable degree,” Wielgus says.

 

Other common missteps include reacting to one-off occurrences versus continuous measurement and ignoring predictable utilization spikes. Teams should document predictable high utilization windows like month-end closes, nightly jobs or seasonal spikes, and create an operational threshold that reflects business risk and the IT team’s performance and uptime commitments.

Monitoring and Alerting: Critical Tools

Real-time signals give teams time to act, not just explain what went wrong later. Start with native tools. In Navigator for i, system monitors can trigger actions and automated responses based on the team’s set thresholds. As Barsness explains, “If utilization goes over a designated threshold, Navigator can send an email.” Other options include building in pre-designated responses based on key thresholds; for example, holding a job queue.

 

Cross-platform visibility matters, too. IBM i shouldn’t live in a monitoring silo—pair native monitors with tools like Instana. Make sure to build tracking and cross-checks into routine reviews to head off minor issues that could become performance incidents. Finally, avoid tunnel vision: As Barsness says, “Many teams monitor what bit them the last time, not what might bite them next.”

Many teams monitor what bit them the last time, not what might bite them next.
—Eric Barsness, principal solution architect and master inventor, IBM

Capacity Planning Missteps and Wins

​Misstep →
Upgrading the wrong thing
A customer added CPU but saw no performance gain. The analysis showed storage was the real bottleneck: “They changed the storage and chose a less costly storage that didn’t perform as well,” says Barsness. The lesson: Confirm the real bottleneck first and size to that—before spending on hardware.
Win →
Benchmark before buying
Teams that simulate real-life system stressors and growth make better decisions. Barsness recommends that teams first do a true scalability test to benchmark, and then “drive the heck out of that thing and model several years of growth.” Surfacing hidden limits early helps teams right-size capacity, storage and configurations.
The Bottom Line
A simple guiding principle is to optimize for resilience. As Wielgus reminds us, “The question is not about being performant; the question is about being safe.” Thoughtful tests plus the right metrics keep IT teams out of firefights and on target.

Putting It All Together

At the end of the day, the playbook is simple: Collect the right data, monitor for the right signals and review in the right cadence.

 

As Barsness puts it, “Performance and capacity planning are important for a number of reasons: The end user experience, consistent good performance and being efficient with your resources.” The payoff for technology professionals who take a proactive, strategic approach? Predictable response, right-sized capacity and fewer surprises.
Share this article
Share this article