Expert Blog: AI and Predictive Analytics – Improving Data Center Lifecycle

Photo Gavin Flynn
Author: Gavin Flynn, Director (US-Central Region), Linesight

Every second, every person online creates 1.7MB of data. Understanding and harnessing all of this information allows us to predict trends, reduce risk and improve how businesses work. However, the sheer volume of information passing through data centers makes it difficult to collect and analyze this data, which impedes performance.

The solution lies in another brainchild of the digital age: artificial intelligence (AI). With the new abilities of AI, data centers can focus on using predictive data to improve efficiencies during construction and throughout the lifecycle of the data center. With this sharpened focus, data centers can reduce downtimes and ensure that valuable assets are monitored and safeguarded.

Challenges Facing Construction Companies

When considering the challenges faced with bringing data center projects to completion, the importance of historical data in supporting accurate budgeting cannot be overstated. Budgeting has been especially difficult since the 2008 recession reduced the construction labor force by 60 percent. Material prices are also a concern, especially as they can vary seasonally based on the commodities market and a project’s construction volume. With materials, labor and equipment accounting for an average of 79 percent of total project costs, all of these factors are vulnerable to a volatile market, resulting in unpredictable data. This in turn can create unreliable budget projections.

This only illustrates the failures of traditional economic forecasting techniques to predict cost volatility and account for sudden and violent market changes, as we’ve seen not just with the Great Recession, but even more recently with the COVID-19 pandemic. Outdated predictive systems simply cannot account for world-spanning diseases. The solution for the data center construction sector is to shift from older models of trend-based analysis to predictive cost data for a more robust, reliable and data-driven alternative.

The Benefits of Predictive Cost Data 

Predictive cost data mixes classical economic prediction models with cutting-edge data mining analytics. This creates highly reactive techniques that are driven by new data. Traditional econometrics, in turn, provide context for the emerging information from data mining. This system has been used to improve accuracy on construction costs up to three years before a project breaks ground, giving developers a long lead time for preconstruction. Although no model can be 100 percent accurate, predictive cost data uses historical trends and is the most accurate tool available.

Data mining also uses the ongoing increases in computing power and statistical analysis procedures to identify patterns and determine when and how construction and labor costs will change. By properly leveraging modern data techniques like these, developers can improve the entire pre-construction phase, supplementing historical data with more accurate projections of future costs that account for potential anomalies. Six to twenty-four months before breaking ground, it becomes possible to maintain an accurate estimate of costs, as long as the data necessary to construct both predictive and descriptive analytics models can be collected.

Benchmarking and Machine Learning

Descriptive analytics, or benchmarking, features moment-to-moment tracking of a company’s output as compared to other leaders in the industry. This creates a complex and nuanced idea of how certain techniques and approaches are performing, leading to better cost intelligence.

Industry experts rely on benchmarking data to create accurate predictions of project cost and scheduling per region, year and project. In conjunction with machine learning, this tool can help with delivery and cost analytics by taking into account potential project delays caused by supply chain disruption and inevitable fluctuations in the cost of materials. The global predictive analytics market is expected to reach $10.95 billion by 2022 and has enjoyed an annual growth of 21 percent since 2016, demonstrating just how profitable this new technology could be for new data center projects.

Improving Operational Efficiency

AI and predictive analytics can also optimize data center operations. For example, AI models can be built to monitor equipment performance and determine new ways to improve efficiency, increasing output while reducing the need for repairs. AI systems can also assist with server optimization, refining storage systems, speeding up processing times and reducing risks. The more data that is fed to the server, the faster the AI can deploy solutions. The technology can isolate and bypass faulty servers, while automating security defenses without relying on workers to do so manually, therefore reducing security overhead.

With carbon emissions and energy consumption such a pressing concern in these times, AI can even improve energy efficiency. For instance, large corporations like Google have reduced cooling system usage by 40 percent, thanks to deep learning technology.

An intelligent path to better data centers

Active data centers are constantly processing large quantities of data, not only in their servers but also through their Building Monitoring Systems (BMS) and Data Center Infrastructure Management (DCIM) software. With BMS and DCIM data, data center facilities can leverage AI and machine learning algorithms to increase operational efficiency. Analytics can lead to faster processing time and better use of machines and cooling technology, resulting in less downtime, and reducing breakdowns and other costly setbacks. Optimized facilities, utilizing advanced analytics to run more intelligently, can maximize operational profitability.

During the pre-construction and construction phases, processes also need to become data-driven to create better results. With a real-time understanding of operational costs, Total Cost of Ownership (TCO) calculations can be completed accurately to ensure correct decisions are being made by the design and construction teams throughout the lifecycle of a data center, despite potential CapEx constraints. Overall, data-driven insights can improve efficiencies both throughout the construction project, and throughout the lifecycle of the active data center.