What Is a Server? Types, How It Works and How Businesses Choose One

Every time an employee opens a shared spreadsheet, sends an email or loads the company website, a server somewhere answers the request. Servers sit behind almost every digital business process, yet for many managers they remain a vague idea: a humming box in a cupboard, or something “in the cloud” the IT provider looks after.

This guide explains what a server is, how it differs from a desktop, which jobs servers perform and how the main hosting options compare. It also covers the numbers that matter in a contract, such as what a 99.9% uptime promise actually allows, and the questions to ask before you buy, rent or migrate.

What exactly is a server?

A server is a computer system that provides data, resources or services to other computers, called clients, over a network. The word describes what the machine does, not what it looks like. A laptop sharing a folder with a colleague is technically a file server; a data center rack of machines does the same job at far greater scale and reliability.

The term covers two layers. There is the server hardware, the physical box with processors, memory and storage, and the server software, the program that listens for requests and responds, such as a web server or database engine (our explainer on hardware vs software covers the distinction). When IT says “the mail server is down,” it may mean the machine has failed, or only that the mail program has stopped.

How does the client-server model work?

In the client-server model, a client such as a browser, email app or point-of-sale terminal sends a request, and the server does the work and returns a response. Both sides follow an agreed protocol: HTTPS for web pages, SMTP and IMAP for email, SMB for Windows file shares.

Here is what happens when a staff member opens an internal sales dashboard:

1. The browser asks a DNS server to translate the dashboard’s name into a network address.
2. It sends an HTTPS request to the web server at that address.
3. The web server passes the request to an application server running the dashboard’s code.
4. The application server queries a database server for this month’s figures.
5. The results travel back and appear on screen, usually in under a second.

Several servers cooperated to answer one click, so one failing can break an apparently unrelated application. Our guide to computer networks explains how those requests travel.

How is server hardware different from a desktop computer?

A server contains the same building blocks as a PC: processors, memory, storage and network ports. The difference is in specification and packaging, because a server is expected to run around the clock for five years or more, serve many users at once, and be repaired without anyone noticing.

ECC memory

Servers use error-correcting code (ECC) RAM, which detects and fixes single-bit memory errors on the fly. On a desktop, a rare flipped bit might crash a program; on a server holding a company database in memory, it could silently corrupt records. ECC needs a compatible processor and motherboard and is standard for machines storing important data. See RAM vs storage for what memory does.

Redundancy

• Dual power supplies, ideally on separate circuits, so one can fail while the other keeps running.
• RAID storage, which mirrors or spreads data across drives so losing one drive loses no data. It does not protect against deletion or ransomware.
• Hot-swappable parts: drives, fans and power supplies replaceable while the server stays on.
• Multiple network ports that can be bonded for bandwidth or failover.

Form factors and remote management

Tower servers resemble a large desktop and suit a small office. Rack servers slide into a standard 19-inch rack, with height measured in “U” (1.75 inches), so a 1U server is very slim and a 4U server holds many drives. Blade servers are thin modules sharing power and cooling in a chassis where density matters most. Most server boards also include a management controller (such as iDRAC, iLO or IPMI) that lets an administrator power-cycle the machine, check hardware health and reinstall the operating system remotely.

What jobs do servers do?

Servers are usually named after the role their software performs. One machine can carry several roles, though separating important ones simplifies troubleshooting and security. Roles run on a server operating system, commonly Linux or Windows Server; see operating systems explained.

• Web server: delivers websites and web applications, using software such as Nginx, Apache or IIS.
• File server: stores shared documents with central permissions and backups. A NAS device is a simplified file server.
• Database server: stores structured data such as customers and invoices and answers application queries, for example SQL Server, PostgreSQL or MySQL (see databases explained).
• Mail server: sends, receives and stores email. Most smaller firms now rent this as a hosted service, because spam filtering and deliverability are demanding.
• Application server: runs ERP, CRM or custom software between users and the database.
• DNS server: translates names into IP addresses; when it fails, many other services appear to fail too.
• Print server: manages queues and drivers for shared printers.
• Directory server: holds user accounts and permissions, such as Active Directory, so staff log in once.

What is the difference between a physical and a virtual server?

A physical server, or bare metal, is one machine running one operating system. A virtual server, or virtual machine (VM), is a software-defined computer running on a physical host alongside other VMs. A hypervisor such as VMware ESXi, Microsoft Hyper-V or KVM divides the host’s processors, memory and storage among the VMs and keeps them isolated.

Virtualization became the default because typical physical servers sat mostly idle. Consolidating ten lightly used machines onto one or two hosts cuts hardware, power and space costs, and VMs can be snapshotted before risky changes or moved between hosts, which eases maintenance and disaster recovery. Bare metal still suits workloads needing a machine’s full, predictable power, specialized hardware such as GPUs, or core-based licensing. Containers, a lighter form of isolation, are covered in virtualization and containers.

Where can a business’s servers live: on-premises, colocation or cloud?

On-premises means you buy the hardware and install it in your own building. You control everything, but you also handle power, cooling, physical security, connectivity, spare parts and replacement.

Colocation means you own the hardware but rent rack space, power, cooling and connectivity in a professional data center. You gain generator-backed power and redundant internet links without building a server room, but someone still maintains the machines, and hands-on repairs need a site visit or the facility’s paid “remote hands” service.

Cloud means you rent capacity and never touch the hardware:

• VPS (virtual private server): a fixed-size VM at a predictable monthly price, popular for websites and small applications.
• Dedicated server: an entire physical machine rented from a hosting company.
• IaaS (infrastructure as a service): on-demand VMs, storage and networks from providers such as AWS, Microsoft Azure and Google Cloud, billed by usage and scalable in minutes.

Many companies end up hybrid, for example SaaS email, customer apps on IaaS and a small on-site server for local equipment.

What do uptime and SLAs really mean?

Uptime is the share of time a service is available, and providers promise a level of it in a service level agreement (SLA). The percentages look similar, but the permitted downtime is not:

• 99%: about 3.65 days per year, or roughly 7.3 hours a month.
• 99.9% (“three nines”): about 8.76 hours per year, or roughly 43.8 minutes a month.
• 99.99% (“four nines”): about 52.6 minutes per year, or roughly 4.4 minutes a month.
• 99.999% (“five nines”): about 5.3 minutes per year.

The math: a year has 8,760 hours, so 0.1% is 8.76 hours and 0.01% is about 52.6 minutes. Each extra nine cuts allowed downtime tenfold and usually raises cost sharply, because it requires duplicate servers, automatic failover and often multiple locations.

Read the fine print. Most SLAs exclude scheduled maintenance, and the remedy for a breach is typically a small service credit, not compensation for lost sales. A cloud SLA covers only the provider’s infrastructure, not your own application. For an internal file server, 99.9% may be fine; for an online checkout, an hour offline may cost more than a year of extra redundancy.

How are servers kept secure and maintained?

A server concentrates data and access in one place, which makes it a prime target. Keeping it healthy is ongoing work:

• Patching: prompt operating system, firmware and application updates, since many breaches exploit already-fixed flaws.
• Access control: individual admin accounts, multi-factor authentication and least privilege.
• Network protection: firewalls exposing only necessary ports, and VPN or zero-trust access instead of open remote-desktop ports.
• Backups: automated and tested, following the “3-2-1” rule (three copies, two media types, one off-site), with one copy immutable or offline.
• Monitoring: alerts for disk space, failing drives, high load, failed logins and expiring certificates.
• Lifecycle: hardware refreshed roughly every four to six years, and operating systems upgraded before support ends.

Moving to the cloud shifts hardware and facility security to the provider, but under the shared responsibility model you still configure access, patch your systems and protect your data. See our introduction to cybersecurity.

How should you think about the total cost of ownership?

Comparing a server’s purchase price with a monthly cloud bill is misleading. Total cost of ownership (TCO) counts everything spent to run a service over three to five years:

1. Hardware or subscription: the server itself, or the hosting fee.
2. Licenses: operating system, database, virtualization and backup software, sometimes priced per core or per user.
3. Facilities: electricity for the server and its cooling, a UPS, rack space and physical security.
4. Connectivity: business internet lines and, in the cloud, egress charges when data leaves the provider.
5. People: hours spent installing, patching, monitoring and fixing, often the largest and most underestimated item.
6. Downtime risk: the hourly cost of an outage multiplied by its likelihood under each option.

As a rough pattern, owned hardware tends to be cheaper for steady workloads at high utilization over many years, while cloud tends to win for variable, growing or short-lived workloads and for firms without IT staff. A cloud VM left running at full size around the clock can cost more over five years than an owned server; the same VM resized to real demand may cost far less.

When does a small business actually need a server?

Many small businesses need no server of their own. Email, file sharing and collaboration come from suites such as Microsoft 365 or Google Workspace, and accounting, CRM and HR tools are usually SaaS. The servers exist, but they belong to your vendors. Your own server, on-site or rented, is worth considering when:

• Key software, such as some industry ERP, CAD or practice-management systems, must be installed on your own server.
• You work with very large files, such as video or engineering drawings, that sync slowly over your connection.
• On-site equipment, such as production machines, cameras or tills, must keep working during an internet outage.
• Contracts or regulations require data to stay in a specific location or under your direct control.
• Your website or customer portal has outgrown shared hosting, where a VPS is usually the first step.

What does this mean for your business?

Start from services, not hardware. List each function, who uses it and from where, and what an hour of downtime would cost. That list sets the uptime you need, which drives architecture and budget.

• SaaS first: if a reputable cloud service does the job, it usually removes a server entirely.
• Hosting model: for anything you must run yourself, compare on-premises, colocation and cloud on five-year TCO. Our guide on how to choose a cloud provider compares the major platforms.
• Backups: budget for them from day one; our backup and recovery software comparison is a good starting point.
• Management: decide who patches and monitors. Many small firms use a managed service provider running remote monitoring tools; see our RMM software comparison.

Questions to ask a vendor or IT provider:

1. What uptime is guaranteed, what is excluded, and what is the remedy if it is missed?
2. Where is our data stored, how is it backed up, and when was a restore last tested?
3. Who patches the operating system and applications?
4. If a drive, power supply or the whole server fails, how long until we are running again?
5. What are all recurring costs, including licenses, bandwidth, egress and support?
6. How easily can we move our data elsewhere if we leave?

Frequently Asked Questions