Comprehensive Overview of Data Centers: Definitions, Types, Technologies, and Trends

Data Center (DC) refers to the infrastructure where an organization concentrates computer systems and related equipment such as communication and storage devices, or it can be an outsourced location for multiple companies to store their equipment or data.

Internet Data Center (IDC) is a service platform equipped with high‑speed internet access, high‑performance LAN, reliable power and cooling, professional management, and comprehensive applications, distinguished from a generic data center by its network connectivity and external service provision.

Cloud‑computing data centers are designed to support cloud workloads by abstracting hardware and software resources into services, offering flexibility, elasticity, and scalability. Their characteristics include:

Typically associated with large‑scale data centers, making detailed optimizations significant.

All resources are abstracted as services, enabling flexible and scalable operations.

Accelerating the evolution of underlying infrastructure technologies.

Data center scale is classified by rack count: small/medium (< 3000), large (3000 ≤ n < 10000), and hyperscale (n ≥ 10000). Availability levels follow the GB50147‑2017 standard (A, B, C) or the TIA‑942 standard (T1‑T4), while some vendors use non‑standard “star” ratings.

The lifecycle of a data center consists of planning, design, construction, and operation, each involving distinct upstream and downstream industry chains; the construction phase is the most extensive, covering land, power, water, networking, IT and non‑IT equipment, civil works, and software systems.

From a long‑term perspective, data centers and cloud computing remain competitive: many small‑scale customers are shifting from traditional IDC to public cloud, and as public‑cloud services improve, larger customers may also prefer cloud solutions, pressuring traditional IDC market share.

Domestic IDC providers are actively deploying private, hybrid, and cloud MSP solutions using OpenStack or Kubernetes, and the broader definition of IDC now encompasses cloud computing services.

Data center energy consumption consists of four parts: IT equipment, cooling systems, power distribution, and lighting/others. The Power Usage Effectiveness (PUE) metric compares total energy to IT‑only energy; regional climate, electricity pricing, and regulatory PUE requirements affect overall efficiency.

Considering power constraints, expanding data centers to peripheral or remote regions is a growing trend, though traditional IDC faces challenges in customer onboarding, operations, and inter‑data‑center connectivity, especially for government clients.

From a supply‑side view, transmission latency includes link latency and node latency, which depends on node count. Business sensitivity to latency dictates placement: high‑sensitivity workloads favor core‑city or edge data centers, while medium/low‑sensitivity workloads can be hosted in large, remote facilities to reduce cost.

Compared with UPS, High‑Voltage Direct Current (HVDC) offers superior backup capability, scalability, and lower investment and operational costs, resulting in higher efficiency and smaller footprint.

Cooling systems dominate non‑IT energy consumption. Emerging liquid‑cooling technologies, especially immersion cooling, can reduce PUE below 1.2, and combined with other techniques may approach a PUE of 1. However, adoption is limited by application suitability, coolant cost, and retrofit expenses; increasing GPU workloads and server density will drive broader liquid‑cooling adoption.

Traditional hierarchical network architectures suffer from bandwidth bottlenecks at the root. To address growing east‑west traffic, fat‑tree designs were introduced, but they still encounter congestion under heavy loads. The leaf‑spine architecture, a flatter, more mesh‑like topology using ECMP for dynamic multi‑path selection, improves bandwidth utilization, predictability, scalability, security, and availability.

The leaf‑spine shift has two major impacts: a surge in optical module demand due to full‑fiber deployment, and the practical realization of SDN, which had long been conceptually proposed.

The data center industry is entering a consolidation phase. Urban restrictions and financial instruments like REITs pressure older, smaller facilities, leading to horizontal mergers. Simultaneously, public‑cloud growth drives IDC operators toward private‑cloud, hybrid‑cloud, and deeper integration with upstream vendors developing chips, white‑box switches, and power solutions.

Source: China Data Center Industry Analysis (with Report)