Understanding Servers: Definitions, Architecture, Components, Classifications, and the X86 vs ARM Debate

Introduction Servers are the core infrastructure for cloud computing and data centers. In the context of expanding public cloud services, the server industry is reaching a turning point. This article examines four key questions to give a deep analysis of servers.

1. What is a server? A server (English: "Server") is a high‑performance computer that provides various services over a network. It stores and processes about 80% of network data, earning the nickname "the soul of the network". Compared with ordinary PCs, servers demand higher stability, security, and performance, leading to differences in CPU, chipset, memory, storage, and networking hardware.

Key differences include:

One‑to‑many communication: Clients such as PCs, tablets, and phones access services through the server.

Resource sharing: The server listens for client requests and shares attached disks, printers, modems, and other devices.

Stronger hardware: Higher computational speed, continuous reliability, and greater I/O throughput.

Servers are designed for specific scenarios such as file exchange, data storage/query, and application execution.

2. Server composition

2.1 Logical architecture The logical architecture resembles that of a PC but with stricter requirements for performance, stability, reliability, security, scalability, and manageability. The CPU and memory are the most critical components.

2.2 Hardware Main hardware components include processor, memory, chipset, I/O (RAID cards, network cards, HBA cards), storage drives, and chassis (power supply, fans). Cost breakdown (typical server): CPU & chipset ~50%, memory ~15%, external storage ~10%, other hardware ~25%.

2.3 Firmware and OS

BIOS/UEFI : The first software executed after power‑on, handling hardware initialization and bootstrapping.

BMC (Baseboard Management Controller) : Enables out‑of‑band monitoring, firmware updates, and management even when the server is powered off.

CMOS : Stores BIOS settings in a small RAM chip on the motherboard.

Operating System : Manages hardware and data resources; available in 32‑bit and 64‑bit versions, with 64‑bit offering greater memory and application handling.

3. Server classification

3.1 By form factor

Tower server : Stand‑alone chassis, low density, typically single‑processor.

Rack server : 19‑inch width, height measured in “U”, optimized for space efficiency in data centers.

Blade server : Multiple server blades share a common chassis, offering high density.

Cabinet server : Integrated compute, network, and storage modules, designed for modern data‑center deployments.

3.2 By instruction‑set architecture

CISC (X86) : Intel/AMD processors, dominant in the market.

RISC : Includes IBM Power, SPARC, and Huawei Kunpeng (ARM‑based) processors.

EPIC : Intel Itanium and similar designs.

3.3 By processor count Servers are categorized as single‑socket, dual‑socket, quad‑socket, etc. Dual‑socket servers are currently the mainstream, employing Symmetrical Multi‑Processing (SMP) to share memory and I/O across CPUs.

3.4 By application type

File server : Provides shared file storage for LAN users.

Database server : Handles frequent read/write and indexing operations.

Application server : Hosts and runs business applications for multiple users.

4. X86 vs ARM debate

Servers can be divided into CISC (X86) and non‑X86 (RISC/EPIC) groups. The market is trending toward a dual‑strength situation: X86 servers dominate with Intel/AMD, while ARM‑based servers are gaining traction for specific workloads. In China, the “new infrastructure” policy is expected to boost X86 server demand, with a projected CAGR of 9.1% (2020‑2024). Intel’s upcoming 10 nm Ice Lake‑SP processors (up to 38 cores, 76 threads, PCIe 4.0) may further stimulate X86 server growth.

Recommended reading : Advanced Architecture Design for Distributed Concurrency .