Muhammed Baki SUMER

Kategori: Data Center

  • From Data to Power: The Hidden Architecture of Artificial Intelligence

    From Data to Power: The Hidden Architecture of Artificial Intelligence

    Today, artificial intelligence has quietly become a natural part of everyday life. When we wake up to a phone alarm, choose the fastest route in traffic, or scroll through content recommendations on digital platforms, AI-powered systems are working in the background. But are we truly aware of it? How often do we stop and think about how many times we interact with AI in a single day?

    AI refers to systems that can learn from data, analyze information, and improve over time. I often use this example: AI is like a modern library, and each AI program is like a librarian who finds exactly what you’re looking for. Thirty years ago, we relied on encyclopedias. Then we began using search engines like Google to find information. Today, we use AI programs. Over time, the time required to access information has decreased dramatically.

    Unlike traditional software, which operates based on predefined rules, AI systems learn from vast datasets and generate faster and more accurate results. This raises an important question: what truly makes artificial intelligence powerful—the algorithms themselves, or the data that feeds them?

    Today, AI is actively used in many areas, from social media and banking to navigation and enterprise applications. However, behind this widespread usage lies a massive infrastructure that is often overlooked. Why does artificial intelligence require so many resources? Why does every new AI application increase the demand for data center capacity?

    In the digital ecosystem, data can be considered the brain, while data centers function as the muscles. Just as the brain requires muscles to transform decisions into action, data requires data centers to process, transmit, and operationalize information. Without robust and diverse data centers, sustainable data flow would not be possible.

    The answer to these questions lies in global fiber connectivity infrastructure. Fiber networks spanning the world form the backbone of the internet and enable seamless global communication. At the same time, this infrastructure provides access to enormous pools of data accumulated worldwide, making it possible to reach information, applications, and systems quickly and securely without geographical limitations. But here is the critical point: are these networks used only by humans?

    Today, machines, sensors, applications, and systems are constantly communicating with one another. As a result, data is generated everywhere, transferred continuously, and often processed in real time. As real-time data processing becomes more widespread, why do speed and low latency become so critical? And when we talk about speed, are we referring only to network performance—or something much broader?

    In reality, speed also means high-capacity IT infrastructure, powerful servers, GPUs, and AI accelerators. These technologies are essential for training AI models and running AI applications efficiently. However, as computing power increases, electricity consumption and heat generation inevitably rise as well. How prepared are traditional data centers to handle this growing load?

    Increasing computational capacity is not only a technical issue; it is also a major energy planning challenge. As AI systems scale, the energy demand of data centers grows exponentially. At this point, not only energy efficiency but also the sustainability and continuity of energy supply become critical.

    Today, renewable energy sources offer important solutions for data centers. However, in the future—particularly for high-density, large-scale AI clusters—facility-based energy generation models may become more prominent. In this context, integrated nuclear energy solutions such as Small Modular Reactors (SMRs) could be evaluated as a potential alternative to meet the increasing and uninterrupted energy demands of data centers.

    Facility-based nuclear energy production, particularly through uranium-fueled systems, offers high energy density and continuous power generation capacity. Considering that AI infrastructures are required to operate 24/7 without interruption, energy continuity becomes a strategic necessity.

    These developments demonstrate that data centers are being redesigned not only in terms of digital infrastructure but also in terms of energy architecture. In the future, data centers may evolve beyond simple processing facilities into integrated technology campuses capable of generating their own energy.

    At this point, it becomes clearer why data centers are evolving so rapidly. Increasing computational demands have made conventional designs insufficient. As a result, liquid cooling systems, direct-to-chip cooling solutions, UPS systems designed for high-density environments, high-capacity intelligent PDUs, and open-frame high-density rack designs are becoming increasingly common in modern data centers. When all these technologies come together, what emerges is no longer a simple data center, but a complex ecosystem where power, cooling, networking, and IT infrastructure must operate in perfect harmony.

    Although AI applications appear to users as software or digital services, behind them lie extremely powerful physical infrastructures. Without these infrastructures, could artificial intelligence truly function?

    At this stage, artificial intelligence can no longer be considered just a software development topic. It represents an end-to-end infrastructure transformation—from data generation and network connectivity to computing power, energy management, and cooling systems. As more data is produced, more systems become interconnected, and greater computational power is required, where will this transformation lead?

    Perhaps the real question is this: is artificial intelligence transforming data centers, or is the evolution of data centers what truly makes artificial intelligence possible?

    I will reach out to another points in my upcoming blog posts. Wishing you a smooth path ahead—may everything in your life run as smoothly and beautifully as a well-designed data center. Here’s to success and happiness!

  • Optimizing Data Center Cooling with Compressors

    Optimizing Data Center Cooling with Compressors

    Today, I want to talk a bit about compressor types used in data center cooling systems and how the differences between them can affect overall design. It’s also important to understand which compressor type fits best with which kind of system layout or capacity.

    Let’s start by understanding how these compressors actually work. Scroll compressors operate using two spiral-shaped scrolls—one fixed and one that moves in an orbiting motion. This design compresses the refrigerant in a smooth and quiet way. Because of their simple and compact structure, they are often used in smaller systems or modular solutions.

    Screw compressors work differently. They use two interlocking helical screws to compress the refrigerant in a continuous, smooth flow. This type of compressor is more suitable for higher loads and is commonly found in industrial HVAC systems and large-scale chillers. The operation is stable, reliable, and much more efficient when the cooling demand is high.

    As I mentioned, the most efficient compressor depends on the system design. For example, if your data center has relatively low capacity, scroll compressors are usually a better fit. They perform better under partial loads and can handle frequent start-stop cycles without issues. So if the cooling system is expected to turn on and off often or run at variable loads, scroll compressors offer a solid, efficient choice.

    However, if you’re designing a high-capacity data center, the better option is definitely the screw compressor. In fact, it’s quite rare to find a high-capacity chiller that uses scroll compressors—most of them come equipped with screw compressors by default. That’s because the design and working principles of screw compressors are better suited to high cooling loads.

    From my experience, I’ve seen cases where end users chose in-row or large quantities of DX in-room cooling systems with scroll compressors for high-capacity data centers. This kind of decision often leads to problems later. I’ll go into more detail on that topic in another post, but to stay focused here: in-row and DX in-room units typically use scroll compressors, and they are really only intended for small or medium-scale environments. When used in larger systems, they can cause multiple issues—like service disruptions, lower energy efficiency, and poor cooling performance overall.

    Luckily, this kind of mismatch doesn’t happen very often in professional data center projects. In most large-scale setups, the cooling system is based on chilled water chillers that use screw compressors. These compressors offer more stable operation thanks to their continuous compression cycle rather than a pulse-based approach. This leads to greater efficiency, especially when cooling demand is high.

    When screw compressors are combined with variable speed drive (VSD) systems, energy consumption is optimized even further. They adjust their speed according to the system’s needs, which reduces wasted energy and improves performance per ton of cooling.

    Of course, there’s always an exception. In systems where frequent start-stop cycles are required—like smaller data centers or modular edge setups—scroll compressors might actually be more efficient. Their ability to handle cycling without losing performance makes them a valuable option in those cases. modular edge setups—scroll compressors might actually be more efficient. Their ability to handle cycling without losing performance makes them a valuable option in those cases.

    I will reach out to another design points in my upcoming blog posts. Wishing you a smooth path ahead—may everything in your life run as smoothly and beautifully as a well-designed data center. Here’s to success and happiness!

  • The Golden Guide to Data Center Design and Operations: Why ASHRAE Matters

    The Golden Guide to Data Center Design and Operations: Why ASHRAE Matters

    Hello,

    Today, I want to talk about ASHRAE and why it is important for data center design. First of all, ASHRAE (the American Society of Heating, Refrigerating and Air-Conditioning Engineers), as you can understand from the name, is an essential authority on cooling analysis for data centers. This standard (ASHRAE TC 9.9) defines the ideal cooling conditions for IT equipment. ASHRAE includes the A1, A2, A3, and A4 classifications, which specify recommended operating temperature and humidity ranges. Another important aspect covered by ASHRAE is dew point management.

    ASHRAE is not widely known by the general public when it comes to data center design and operation. However, ASHRAE directly affects your PUE and overall efficiency. According to ASHRAE, your IT equipment’s inlet air temperature should be maintained between 18 and 27 °C. This is a wide range, and system efficiency is directly related to discharge air temperatures.

    If I design based on 18 °C, it keeps the system safe, but it is not efficient. However, if I design for 27 °C, it means achieving the best cooling efficiency and lower energy consumption. Depending on your choice, your system could be up to 35% more efficient.

    What is the impact on the cooling system? Actually, higher discharge air temperatures mean:

    • Lower chiller compressor consumption
    • Lower in-room unit fan speeds
    • Reduced dehumidification requirements
    • Longer free cooling periods

    These features directly improve the overall system efficiency. And efficiency also means sustainability. With this approach, you are essentially determining the system efficiency during the design phase. This is only one of many design considerations for data centers, but it is one of the most important ones.

    I will reach out to another design points in my upcoming blog posts. Wishing you a smooth path ahead—may everything in your life run as smoothly and beautifully as a well-designed data center. Here’s to success and happiness!

  • Have you ever heard of Tier 5?

    Have you ever heard of Tier 5?

    Hello,

    Let’s take a look at the biggest data center in the world today. If you have any comments or additional information, please feel free to share your thoughts in the comments section.

    Yes, you heard that right. I’m talking about Tier 5. Of course, this is not an official classification by Uptime Institute. Instead, it’s a slogan used by Switch Data Center Campus. Let’s explore together why they refer to their data center system as Tier 5.

    Technically, their data center meets the criteria for Tier 4, but Switch uses the term “Tier 5” as a premium quality designation to highlight their enhanced infrastructure security, commitment to green energy, and comprehensive redundancy strategy.

    Their total capacity is 650 MW, and each rack supports more than 55 kW of power. The campus covers nearly 7 million gross sq FT and is powered by solar, wind, and geothermal energy sources. The data center consists of independently operated modules that do not affect each other in case of failure.

    They have on-site fire and security teams, making it feel like a small city. Each cooling infrastructure system is fully separated, just like all other critical infrastructures. Encapsulated structures are used to isolate equipment from fire, flooding, and other disasters.

    Not only power and cooling but also physical security, cabling, and network infrastructure are all fully redundant. A 10-layer physical security system is in place, including biometrics, armed security, and zoned access control. The facility has also been recognized by Greenpeace as the most environmentally friendly data center in the world.

    In today’s post, I wanted to share my thoughts on Switch’s data center infrastructure and design approach, which caught my attention. If you would like to learn more, feel free to visit their website.

    Wishing you a smooth journey—may everything in your life run as smoothly and efficiently as a well-designed data center. Here’s to success and happiness!