Introduction to Servers
Almost every service you have studied so far — DNS, DHCP, web, email, files — is provided by a server. But what is a server, and why does it look different from the laptop on your desk? This lesson defines the server, contrasts it with a desktop, and explains the hardware choices that let a machine run unattended, twenty-four hours a day, for years.
What is a server?
You met the word “server” in Week 1 as a role: the host that provides a service while a client requests it. That role is still true, but when people talk about “a server” in a data centre they mean a specific kind of machine built to play that role permanently. A server is both hardware and software working together.
The hardware is the physical box: CPUs, memory, disks, and network adapters. The software is the operating system plus the service that clients actually use — a web server, a file server, a database engine. Neither half is useful alone: a database program on a powered-off box serves nobody, and a powerful box with no service installed is just expensive metal. In the client–server model the server centralises a resource so that many clients can share it, and one administrator can manage it. That centralisation is the whole point: patch one machine, and every client benefits.
Server versus desktop
On the surface both are computers. The differences are about purpose, and each difference follows from the fact that a server is shared by many users and must not stop.
| Attribute | Desktop / workstation | Server |
|---|---|---|
| Power state | Powered off overnight; reboots for updates | Always on, 24×7, minimal downtime |
| User interface | Local monitor, keyboard, desktop OS | Headless — no screen; managed remotely |
| CPU | One CPU, few cores, built for bursty tasks | Multiple cores/sockets, built for many concurrent requests |
| Memory | Modest RAM | Large RAM with ECC error correction |
| Storage | One or two disks | RAID arrays, often hot-swap drive bays |
| Power | Single power supply | Redundant (dual) power supplies |
| Network | One NIC | Multiple NICs, often teamed for bandwidth and failover |
| Placement | Desk or lap | Data centre, rack, or server room |
Always-on and headless are the two ideas to memorise. Always-on means the machine must survive hardware faults without a human present, which is why redundancy features exist. Headless means there is no local screen or keyboard — administrators connect over the network, typically by remote desktop or a secure shell, so the network connection itself is part of the management path.
Form factors: tower, rack, blade
A form factor is the physical shape and mounting style of the chassis. The three you must know are the tower, the rack-mount, and the blade.
| Form factor | Description | Typical use |
|---|---|---|
| Tower | A standalone upright “pedestal” case, like a large desktop PC. No rack required. | Small business, branch office, first server |
| Rack-mount (1U / 2U) | A flat chassis that bolts into a standard 19-inch rack. Height is measured in rack units (U): 1U = 1.75 in = 44.45 mm. A 2U server is twice that height. | Data centres; high density, easy stacking |
| Blade | A thin server module that slides into a blade enclosure (chassis) which shares power, cooling, and networking across all blades. | Very high-density data-centre deployments |
The rack unit is worth remembering exactly, because exams and data-centre plans quote it: 1U is 1.75 inches, or 44.45 millimetres, tall. A typical full rack is 42U, so you can also think of rack space as a budget you spend one U at a time.
Server hardware worth knowing
Server hardware is chosen for reliability and for handling concurrent work. Walk through the list, because each item answers a specific failure or load problem.
- Multiple CPU cores. A single physical CPU package (a socket) contains many cores, each able to run threads. Servers often have two or more sockets, so dozens of threads can serve many clients at once.
- ECC RAM. Error-Correcting Code memory detects and corrects single-bit errors caused by electrical interference. A cosmic ray flipping one bit in a desktop might crash a game; in a database server it could corrupt a record. ECC prevents that.
- RAID. Multiple physical drives are combined into one logical volume for speed, redundancy, or both — covered in the next section.
- Redundant power supplies. Two power supplies feed the server from different circuits. If one supply or one circuit fails, the server keeps running. They are usually hot-swap so the failed unit can be replaced without powering down.
- Hot-swap drives. Drive bays allow a failed disk to be pulled out and replaced while the server is still running, so the array rebuilds without downtime.
- Dual / teamed NICs. Two or more network adapters are bonded into a team (also called a bond or LAG). This adds bandwidth and, more importantly, failover: if one link dies, traffic continues on the other.
RAID: redundancy from many drives
RAID (Redundant Array of Independent Disks) treats several physical drives as one logical drive. Each level makes a different trade-off between performance, usable capacity, and fault tolerance. Learn the four levels below cold — the numbers and the minimum drive counts are exam favourites.
| Level | Technique | Min. drives | Fault tolerance | Usable capacity |
|---|---|---|---|---|
| RAID 0 | Striping — data is split across all drives for speed | 2 | None (one drive fails, all data is lost) | 100% (N drives) |
| RAID 1 | Mirroring — every drive holds an identical copy | 2 | Tolerates 1 drive failure | 50% (N/2) |
| RAID 5 | Striping with distributed parity — parity spread across all drives | 3 | Tolerates 1 drive failure | (N−1) of N drives |
| RAID 10 | Mirrored stripes (RAID 1 + 0) — pairs are mirrored, then striped | 4 | Tolerates 1 drive per mirror pair | 50% |
Common server roles
A single powerful server can run several roles at once, or each role can get its own machine. Match each role to the service it provides.
| Role | What it provides | Example service |
|---|---|---|
| File server | Central storage for shared documents and folders | SMB/CIFS, NFS |
| Print server | Queues and shares printers to many clients | Windows Print Spooler |
| Web server | Delivers web pages and web applications | HTTP/HTTPS (Apache, Nginx, IIS) |
| Mail server | Sends, receives, and stores email | SMTP, IMAP, POP3 |
| DNS server | Resolves names such as example.com to IP addresses | BIND, Microsoft DNS |
| DHCP server | Hands out IP addresses and network settings automatically | ISC DHCP, Microsoft DHCP |
| Database server | Stores and queries structured data for applications | MySQL, PostgreSQL, SQL Server |
| Application server | Runs the business application logic clients connect to | Java/.NET application tier |
| Virtualisation host | Runs many virtual machines on one physical box | VMware ESXi, Hyper-V, KVM |
Notice that the virtualisation host role is different in kind: instead of serving clients directly, it serves other servers. Lesson 2 covers hypervisors and how one physical server can host many virtual ones.
The rack and the clients
The first figure shows a typical rack elevation. Each rectangle is a server or switch occupying a number of U of vertical space; a top-of-rack switch aggregates their network links.
The second figure steps back from the hardware: a server is only useful because clients reach it across the network to use its services.
Review
Check what you have learned with these review questions.
Sources & references
- Cisco. CCNA (200-301) Exam Topics — network components and services. cisco.com.
- Cisco Networking Academy. Networking Basics (server and client roles, converged networks). netacad.com.
- Microsoft. Windows Server documentation: Storage Spaces and RAID levels. learn.microsoft.com.
- Braden, R. RFC 1122: Requirements for Internet Hosts — Communication Layers (host roles and terminology). rfc-editor.org/rfc/rfc1122.
- IANA. Service Name and Transport Protocol Port Number Registry. iana.org.
Target exams: CCNA 200-301 v1.1 · CyberOps Associate 200-201. Last content verification: 2026-10-11.