Network Engineer’s Guide to Servers: Types, Uses, Network Design and More

In this guide, we’re going to demystify one of the most elementary components of any enterprise network – servers.

What Is a Server?

A server is a computer program or device that provides a service to another computer program and its user, also known as the client. The term server can be used to describe both the hardware typically used to build server devices, and the operating-system software that delivers the service to the clients.

Although servers are designed to carry out a specific function, they’re not that different from other computers. They’re usually built with the same components (except for the operating system (OS) software, which will usually be function-specific).

Server hardware components include the following.

• Motherboard

• Central processing unit (CPU) – most servers use X86/X64 CPUs

• Random Access Memory (RAM)

• Disk space

Server OS software can include the following.

• Windows

• Linux

• Red Hat Enterprise

• Unix

• macOS

As you can see, most of what makes up a server is identical to the desktop or laptop you may be using to read this article.

What’s The Difference Between a Physical and a Virtual Server?

We live in an age of wonder, where rapid technological advances are beginning to outstrip our ability to understand them. So, it’s kind of reassuring that a “virtual” server relies on the same set of components listed above and plugs into the wall socket just like any other computer.

The main difference between a physical and a virtual server comes down to who owns the server and what they’re using it for.

Think of a piece of server OS software as a commuter, who is looking to get from one location to another (to fulfill their function and provide their service).

We could build our commuter a motorbike – designed just for them – to transport them to where they want to go, which is how a physical server works.

Or we could buy a train ticket. There’s room enough for them alongside the other commuters, and if our commuter wants privacy and a good seat, we could go as far as booking a private compartment.

In this analogy, the train is a virtual server. Owned and maintained by larger corporations, our ticket allows us the functions of having our own server, without the need for a physical device of our own – which is the “virtual” part. 

Public and private cloud services are the equivalent of the general seating or private compartment on the train – although, thankfully, if you choose to go public, your cloud-based services are unlikely to include a guy manspreading while eating a tuna sandwich.

Which Server Deployment Do I Need?

That’s the trillion-dollar question – whole empires have risen and fallen attempting to provide the answer. To help you figure out what you need, we’ve summarized the three most common server deployment methods.

Physical Server On-Premise

Your machines, running your software, installed at your location. For businesses with multiple locations, the corporate headquarters is typically the preferred spot. On some occasions, though, distribution centers or satellite offices can provide geographical advantages, such as more diverse network routing options or better overall proximity to reduce latency across the wide area network (WAN).

Pros

• Your internal IT team is in complete control. Your team can customize and upgrade the servers, according to changing business needs. Less downtime, support tickets, and general frustration, if you set things up properly.

• Your server budget is under control. If you’ve built it, you own it. No stealth taxes, inflationary price hikes, or sneaky tiered plans that charge you over the odds when your traffic spikes.

• Offline working remains an option. If your servers are on the same site as your staff, it’s possible to keep business-critical applications running on the local area network (LAN), even in the event of an internet outage. For some businesses, this is a game changer.

Cons

• Equipment costs. You’ll have to buy and build your servers – or pay someone to do it.

• Specialist IT staffing costs. All that extra control is only a blessing if you have the in-house tech talent to pull it off. You’ll need skilled staff to manage and maintain your servers – or (again) pay someone else to do it.

• Energy supply costs. Your electricity bill is going to skyrocket (again). You’ll also need a generator, and battery backup. Not to mention sufficient AC.

• HVAC costs. Too cold? Static electricity, condensation, and water damage. Too hot? Your CPU’s going to melt. Just right? Somewhere between 68 and 71°F. All day long. Did we mention this is expensive?

• Security and maintenance costs. Assuming you have the physical space to do this, you still have to carefully manage access to the servers – otherwise you’re leaving yourself open to a whole new world of security risks, on top of your usual network security threats.

• Your internal IT team is in complete control. Redundancy, support, and more are in your team’s control. If you set up your infrastructure with a single point of failure or with obsolete hardware, you’re in the driver’s seat to deal with issues that arise.

Physical Server in a Data Center (a.k.a. colocation)

Your machines, running your software, in a custom-built, rented data center space and “colocated” with other similar machines of other companies. Colocation allows businesses with data-heavy requirements to manage complicated network architecture without the necessity of owning their own data centers.

Pros

• Stay in control of your servers. You’re still holding all the cards when it comes to building, customizing, and maintaining your servers – though you’ll need to set up a “smart hands” or a “remote hands” solution, to make sure data-center staff can resolve problems with your servers on-site. 

• Location, location, location. You can optimize your network to reduce latency and provide routing diversity, by choosing the data center whose location most closely matches your needs.

• Weapons-grade security. You’d struggle to match data-center standards of security at your own location – there’s widely adopted standardization models such as TIA-942 certification that offer tiered levels of additional protection, in accordance with zero-trust network philosophies.

• Hardened power plant and industrial HVAC as standard. That’s what you’re paying for, right? Let the data center handle the additional energy and HVAC requirements (as well as the procurement, maintenance, and monitoring required to run these functions at your own site).

Cons

• CapEx costs. Although you won’t be paying for HVAC, the initial server build costs are still substantial.

• OpEx costs. Data centers offer their facilities at a significant mark-up, and you’ll be paying higher rates for their power than you would for your own. Knowing how to calculate your colocation power needs is essential.

• No internet? No servers. You won’t be able to access your servers – or any of the associated functions – without a dedicated internet connection, so you’ll need to pay for connectivity in addition.

• No instant fixes. The downside of all that extra security is the time and effort required to access and modify your server installation at the data center.

Virtual Server Installed in the Cloud

Their machines (meaning your cloud provider - AWS, GCP, Azure, etc.), running their software, in their location.

Pros

• No CapEx required – no need to purchase any equipment.

• No maintenance means no additional burden or responsibilities for your in-house IT team.

• No additional power hardening, HVAC, or adaptation of location required.

• Easy to upgrade server storage and performance. Just get in touch with your cloud-service provider and they’ll action the upgrade on your behalf.

• Cloud provider app ecosystem. You can leverage your cloud provider’s application ecosystem to get all sorts of cool software running on your servers in a snap.

Cons

• Lack of control. When you’re taking a ride on someone else’s train (server), you’ve got no input or visibility into what happens to the physical equipment used.

• Vulnerable to an even wider range of points of failure. Your data’s getting routed wherever your cloud-service provider chooses. Extra care needs to be taken to ensure redundancy and uptime.

• Latency is likely. You’ll have less choice over which data centers are used, so any extra geographical distance involved as a result is likely to impact performance.

• Cost creep. At small scale, given the lack of initial capex and low per-unit entry pricing, cloud servers are cheaper than on-prem / colocated server deployments. However, as the scale of your deployment increases, costs grown nonlinearly, oftentimes significantly eclipsing the costs of physical server deployments. Andreessen Horowitz wrote a good piece on this phenomenon in 2021.

Different Types of Servers

Here are 10 common functions that a server provides in a business network environment.

AD (Active Directory Server)

A database of users, and the services they’re authorized to access.

DC (Domain Controller Server)

Works in conjunction with the AD server to authenticate domain users, acting as a gatekeeper.

File Server

Usually, a server with ample disk storage space, used to share resources too big for email between users on the same domain or LAN. 

Exchange (mail) Server

The server equivalent of the post office (or the old telephone exchange and operator), this server acts as a waystation for routing emails.

Print Server

Used to organize print requests from multiple users into a centralized print queue.

FTP (File Transfer Protocol)

Like a file server, FTP servers are used to store and share large files. However, FTP is an internet-based protocol, and requires connectivity to function, unlike file servers which operate within the LAN.

Web Server

Responds to HTTP (Hypertext Transfer Protocol) client requests by storing, processing, and delivering website content and web pages to users.

DNS (Domain Name System) Server

DNS is used to translate IP addresses (an unwieldy string of numbers) into friendlier-looking URLs and back again.

APP (Application) Server

Used to deliver business critical applications on the LAN (as opposed to SaaS applications).

DBS (Database) Server

Used to provide business databases on the local network – popular instances include SQL, Oracle, and dBASE.

Servers and Network Design

So, what’s the best way to incorporate your servers into your network? There are a few handy principles that will help guide your network design decisions.

• Deploy your servers as close as physically possible to your users.

• If your users are spread out, then look at distributing the network architecture and server deployment to suit.

• Weigh up your usage of application and database servers – would SaaS alternatives provide a more effective network design?

• Which servers will lead to network outages, should they fail? Identify any single points of failure, and implement redundancy and backup options accordingly. Then test your backups, regularly.

Servers and Security

Even if you’re running physical servers at your own location, it’s still difficult to fully isolate your servers from the internet and external security risks.

There are some bare minimum, best-practice recommendations we can make to help you sleep a little easier.

• Keep all server installations safely behind a perimeter security firewall.

• Network administrators should be the only users accessing the servers.

• Incorporate multi-factor authentication for server access

• Install endpoint protection, adding an additional layer of security for every server.

• Use Data At Rest Encryption (DARE) to make sure all data remains encrypted, even when it’s not being transmitted.

• Regular (preferably nightly) backups to a secondary site, isolated with an air gap.

Choosing the right server deployment, location, and installation type for your business network can be challenging – especially without the right data. Using Lightyear’s telecom procurement operating system, you can easily configure and procure internet connectivity to your server sites (or direct cloud connectivity if that’s your desire), OR even configure and procure an entire data center space.