Every operating system and application ultimately runs on physical hardware. This guide walks through the core components — CPU, memory, storage, networking and the data centre gear behind cloud services — using tabs so you can jump straight to the part you need.
CPU — The Processor
The Central Processing Unit (CPU) executes instructions — the "brain" of a computer. Modern CPUs have multiple cores (independent processing units on one chip) so they can run several tasks in parallel.
- Clock speed (GHz): how many cycles per second a core can execute — higher isn't always better since architecture and core count matter too.
- Cores & threads: more cores mean more parallel tasks; hyper-threading/SMT lets one core handle two threads.
- Cache (L1/L2/L3): small, very fast memory built into the CPU that stores frequently used data, avoiding slower trips to RAM.
- Architecture: x86-64 (Intel/AMD) dominates servers and desktops; ARM (Apple Silicon, AWS Graviton) dominates mobile and is growing in servers for its power efficiency.
For servers, core count and per-core performance both matter — database and virtualization workloads benefit from more cores, while latency-sensitive workloads benefit from higher per-core clock speed.
RAM & Memory
RAM (Random Access Memory) is fast, temporary storage the CPU uses to hold active programs and data. Unlike storage, RAM is wiped when the power is off.
| Type | Typical Use | Notes |
|---|---|---|
| DDR4 | Most desktops/servers (2024 and earlier) | Still widely deployed in production |
| DDR5 | Current desktops and new servers | Higher bandwidth, lower power per bit |
| ECC RAM | Servers, workstations | Detects and corrects memory errors — critical for uptime |
Running out of RAM forces the OS to use slower disk-based swap/page files, which can dramatically slow a system — this is why memory sizing is one of the first things to check when diagnosing a sluggish server.
Storage
| Type | Speed | Best For |
|---|---|---|
| HDD (Hard Disk Drive) | Slowest (mechanical) | Cheap bulk/archival storage |
| SATA SSD | Fast | General-purpose upgrade over HDD |
| NVMe SSD | Fastest (PCIe-based) | Databases, OS drives, latency-sensitive workloads |
RAID Levels (Redundancy)
- RAID 0: Striping for speed — no redundancy, any disk failure loses all data.
- RAID 1: Mirroring — full redundancy, halves usable capacity.
- RAID 5: Striping with parity — tolerates one disk failure, good capacity efficiency.
- RAID 10: Mirrored + striped — strong performance and redundancy, common in production databases.
RAID protects against drive failure, but it is not a backup — it doesn't protect against accidental deletion, ransomware, or site-level disasters.
Network Hardware
- NIC (Network Interface Card): connects a device to a network; servers often run 1/10/25/100 Gbps NICs depending on workload.
- Switches: connect multiple devices within a network, forwarding traffic based on MAC addresses.
- Routers: connect different networks together and direct traffic between them, typically at Layer 3 (IP).
- Firewalls/load balancers: dedicated appliances (or virtual equivalents) that filter traffic and distribute load across servers.
Data Centre Architecture
Cloud and enterprise services run on racks of servers housed in data centres, built around a few core concepts:
- Rack servers / blade servers: standardized 19-inch rack-mounted hardware for density and serviceability.
- Redundant power (PDU + UPS + generators): ensures uptime through power loss.
- Cooling: hot/cold aisle layouts and CRAC/CRAH units keep equipment within safe temperature ranges.
- Availability Zones/Regions: cloud providers spread hardware across physically separate facilities so one site's failure doesn't take down a whole service.
Understanding the physical layer helps explain cloud concepts like availability zones, instance types, and why certain workloads (databases, GPUs) are priced and provisioned differently than others.
Hardware Lifecycle & Refresh
Unlike software, hardware doesn't have a vendor-published "end-of-life support" date in the same sense — but it does wear out and become obsolete. Typical refresh cycles:
| Hardware Type | Typical Refresh Cycle | Why |
|---|---|---|
| Laptops/Desktops | 3–5 years | Battery wear, performance vs. newer software demands |
| Servers | 5–7 years | Warranty expiry, efficiency gains, vendor support contracts ending |
| Network switches/routers | 5–10 years | Firmware/security update support windows |
| SSD/HDD drives | 3–5 years | Wear leveling limits (SSD) or mechanical failure risk (HDD) |
Most enterprise hardware vendors (Dell, HPE, Cisco) publish their own End-of-Service-Life (EOSL) and End-of-Support dates per model — always check the vendor's official lifecycle page before relying on hardware past its supported window, the same way you would for an operating system.
Keep Learning on ITVedas
One of many free guides across 8 IT chapters — all in plain English.
Explore All Chapters →