It’s no secret that data drives everything these days, and an increasing number of companies are turning to platforms like the AVEVA PI System (also known as PI) to collect, store, manage, and visualize critical operational information in real time.

CSE ICON’s customers are no exception, and the solution works. But as one of ICON’s oil and gas customers discovered, systems of this nature don’t exist in a vacuum and need to evolve along with the business.

As the customer’s reliance on the PI system increased, so did the complexity of resulting data management tasks—the exact opposite of what they were expecting. Performance degradation, slow data retrieval, and persistent system bottlenecks were becoming an increasing threat to operational efficiency.

While this is not an ideal situation for most companies, it’s exactly the kind of challenge ICON is designed to take head on. This case study explores how ICON identified and resolved the performance issues that plagued their customer’s PI system and restored peak efficiency, reliability, and resilience across the entire organization.

Challenge: Addressing Performance Bottlenecks in Real-Time Data Systems

The customer’s PI system had become the backbone of operational decision-making, but constant performance issues were becoming increasingly difficult to ignore. Engineers and IT professionals faced sluggish displays, continuous load bars, and frequent timeouts, which severely impacted the speed and accuracy of data-driven decisions.

    Further metrics revealed that average load times had ballooned from a few seconds to over five minutes in some cases.

    Additionally, RAM usage on the PI Vision server was consistently maxing out, indicating massive inefficiencies. This not only frustrated users but also raised concerns about the system’s ability to scale for future data demands.

    Solution: Identify and Eliminate Bottlenecks, Streamline RAM Usage

    Recognizing the severity of the situation, the customer engaged ICON to conduct a thorough assessment of their PI system. The goal was clear: identify the root causes of performance degradation and implement targeted solutions to restore system integrity.

    Initial Assessment

    The first step was to conduct a detailed evaluation and audit of system architecture. From the very beginning, it was noted that the PI Data Archive (data storage) and AF Server (data modeling and analysis) were hosted on the same machine. This setup is far from ideal, since these components are typically kept separate to distribute processing loads more effectively.

    “Most of our clients have these key systems hosted on different machines,” says Nick DePietro, Sr. Systems Engineer, CSE ICON. “When they’re on the same server, the PI Data Archive and AF Server compete for resources. This gets particularly problematic when multiple users demand data from both components at once.”

    Bottleneck identified.

    Bottleneck Analysis

    Further analysis revealed that the PI Data Archive was prioritizing storage over responding to user queries, particularly under heavy workloads. In these high-demand scenarios, the system went into a “write-only mode” where its primary focus was to store incoming data as efficiently as possible. While this ensured data integrity, it meant that user query responses were delayed or worse—skipped entirely. Additionally, the flood of data requests was overwhelming the server, causing critical slowdowns across the board.

    Data Analysis

    To ease this pressure, the customer added RAM to their PI Vision server. While this provided some relief, it was far from a sustainable solution. ICON’s team noticed that RAM usage on the Vision server would gradually increase until it maxed out, indicating that the system was caching excessive amounts of data.

    “A key contributor to this was user oversight,” DePietro says. “Simple things, like leaving multiple tabs open in complex displays can have a major impact on performance.”

    The ICON team also noticed that the size and complexity of these displays were a core problem, because each display was pulling vast amounts of data (much of it irrelevant to user needs) that further overburdened the system.

    Recommendations and Implementation

    After the initial analysis was complete, ICON presented a series of targeted recommendations and solutions:

    1. Component Separation: To distribute workloads more efficiently and realize significant performance gains, the PI Data Archive and AF Server were separated across machines. This separation allowed each component to operate more efficiently and effectively, because they no longer had to complete for resources.
    2. Display Optimization: The ICON team took a hierarchical approach to restructuring the customer’s displays, streamlining them to focus only on necessary data. Instead of loading all data all at once, users were provided with an overview and the ability to drill down into specific details as needed. This not only reduced load times but also made displays simpler and more user-friendly.
    3. Data Analysis Tools: A PowerShell script was developed to convert the out-of-the-box XML statistics file associated with the PI Analysis Service into a more accessible Excel format. This allowed the customer’s IT team to easily analyze and monitor system performance—and quickly identify and address any (and all) emerging issues.
    4. Template Optimization: The top 10–15 Asset Analytics templates that were causing the most lag were located and optimized. By reducing their execution frequency and simplifying calculations, ICON alleviated much of the system’s burden, leading to smoother overall performance.

    Result: Significant Performance Improvement and Operational Efficiency

    ICON’s recommendations resulted in dramatic improvements in system performance.

    Display load times were slashed from several minutes to mere seconds, which greatly enhanced the user experience.

    The separation of the PI Data Archive and AF Server components eliminated all bottlenecks, allowing incoming data and user inquiries to flow freely—without overburdening either system.

    And finally, the optimization of displays and templates reduced RAM usage, which prevented the system from maxing out and causing additional slowdowns.

    “As a result of this approach, the customer’s PI system became more responsive, reliable, and ready to handle future growth and increasing data demands,” DePietro added.

    About CSE ICON

    CSE ICON specializes in optimizing industrial data platforms with a proven track record of resolving complex performance issues in systems like PI. Companies facing similar challenges can benefit from ICON’s expertise in system analysis, optimization, and implementation of best practices. When you partner with ICON, your organization can be absolutely certain data platforms are running at peak performance to enable faster, more informed decisions. For companies like the one discussed in this case study, where real-time data is critical, the solutions provided by ICON were not just beneficial—they were essential. If your organization is experiencing similar performance issues, CSE ICON is here to help you optimize system performance and gain peak operational efficiency.

    Need Optimization?

    To improve the performance of your AVEVA™ PI System™?

    CSE ICON is a professional services company focused on the design, development, and implementation of Operational Technology used in the processing and manufacturing industries. Our mission is to bring people and data together, ‌helping our customers continuously improve and increase profitability.