- Physical or 'on-premise' servers out perform similarly configured cloud instances in lifetime cost-to-own
- Companies with a wide variety of server needs might need to consider a mix of physical & cloud servers
- When compared to similarly configured cloud instances, the break-even cost for an on-premise server is often less than 2 years
Recent research has shown that over 90 percent of companies surveyed use some manner of cloud computing service to manage their business. With numbers like that, it might be easy to assume that the cloud has taken over the server market, and there’s little justification left for companies to own physical, on-premise servers in today’s ultra-connected world. While there’s no doubt that the cloud computing market is growing incredibly fast, is it always the best or even most cost-effective option for a business in 2020?
In this article, we’ll take you through a head-to-head comparison of on-premise vs. cloud servers for a variety of different use-cases and operational considerations. This comparison will provide insight into the ownership, management and overall cost differences between on-premise and cloud servers, and act as a purchasing guide to help ensure that you make the best long-term decision for your business.
What We're Comparing & Common Uses for Servers
There’s no shortage of companies, or related business services, that exist to help facilitate the ownership and maintenance of servers. From so-called private clouds to public clouds, from co-location data centers to 100% virtualization – the variety of services available today is extensive, allowing for businesses to customize how they own and operate their server fleet. While we could write an entire article on the features and benefits of these services available today, for this article we want to focus on the two most popular ways a business can operate a server: On-premise (physical) servers that are owned by the business using them, and cloud server instances that companies rent from cloud service providers like Amazon Web Services (AWS) or Microsoft Azure.
On-premise: Often referred to as on-site or physical servers, this entails owning, housing and maintaining the actual physical equipment. The server itself can come in a variety of forms such as a tower that looks almost identical to the common desktop PC that sits under most office desks. The form could also be a rack, which are shorter and wider and designed to be stacked vertically in a housing. Rack servers are the type most often used in large server farms and data centers. The form-factor of the server notwithstanding, the important concept to understand here is that the business using the server owns the server and is not renting it, or a portion of it (i.e. a server instance) from another company.
Cloud: For this we are referring to publicly available clouds service providers that allow businesses to rent (i.e. pay-as-you-go) an instance on the physical servers that they (the cloud provider) owns, operates and maintains. There are numerous cloud service providers available today, but by far the largest and most dominant at the time of this publication are Amazon Web Services (AWS), Microsoft Azure and Google Cloud. There is also a collection of smaller, more developer-focused cloud providers that pride themselves on ease of use and omit most of the complicated web interfaces found on the aforementioned dominant providers. Examples of these smaller providers are Digital Ocean and Linode. For this article, we will not be comparing these smaller providers nor will we be comparing private clouds, colocation centers or any hybrids of those services.
The first, and probably the most obvious difference between an on-premise and a cloud server is the physical location of the server itself. An on-premise server is physically located somewhere on the premises of the business that owns it. The business’s own IT administrators, system admins or hired third-party service providers will always have access to the physical hardware. Having this type of access to the server means complete control of things like installation, maintenance, and continued upkeep as well as the ability to service any of the hardware components such as the CPU, memory (Ram) and hard drives.
Cloud servers on the other hand are physically located in large data centers that are owned, operated and serviced exclusively by the cloud service provider. While there are certainly examples of very large companies with incredibly high resource demands that have been able to work out special contracts with specific cloud service provider (think Netflix and AWS), it’s very likely that no one from your company will ever have physical access to the actual hardware that your data, or website is hosted on in the cloud. With that in mind, where your physical server is actually located is solely dependent on the locations of the server farms that are operated by your cloud provider. These days, most of the larger cloud providers do offer their users the ability to chose the location of their cloud server instance from a list of available data centers that are owned by that provider.
Location Winner: Tie
How important the physical location of a server is will be entirely dependent on how your business plans on using it. Does your company deal with highly sensitive or confidential business information? If so, you may be taking a serious risk by putting that information on a publicly available cloud network, and would likely be much better off opting for an on-premise server. Conversely, if your business simply needs a server for hosting a publicly available website – a cloud server instance could be a very easy and suitable option.
Comparison: Configuration Options
When a physical server is purchased, regardless if it’s being bought new from the original equipment manufacturer (OEM) like Dell or HP, or refurbished from an authorized reseller like SourceTech Systems, the buyer is able to completely customize the configuration of the server’s hardware to the exact specifications that he or she needs. From form-factor, which can be a traditional PC-like tower or the more stackable style rack, to the internals such as CPU (including clock-speed and cores), memory (Ram) and hard drive (HDD & SSD) – you can build out your physical server to exact specifications that is required to run your business.
When you build a server in the cloud, what you’re actually doing is creating an instance of a virtual private server (VPS). A VPS allows companies like Amazon and Microsoft to partition (split up) a single server using virtualization software, which allows for dozens of separate, and potentially unrelated, companies to all utilize the same physical server at the same time. The virtualization required to run these VPS instances is built on sophisticated software that’s often proprietary to the specific cloud provider. While not the norm, many cloud providers now offer dedicated virtual private servers, which are different from standard virtual private servers in that they are not shared with other companies. Dedicated private servers however are often much more expensive then a similarly configured virtual private server.
|Firmware / Control Panel OpenManage, EMC||Yes||No
|Operating System Linux, WindowsOS||Yes||Yes
|Hardware Specifications CPU, Memory, Disk Space||Yes||Yes
|Digital Access Limit by IP address||Yes||Yes
|Physical Access Connect directly to device||Yes||No
|Dedicated Equipment One user per server||Yes||No
Configuration Options Winner: On-premise
For companies that want complete control over how their server is configured, including all of the internals like CPU, memory and hard drive, than owning an on-premise server is going to be the best option. Unlike cloud servers, an on-premise server allows for direct physical access to modify, maintain or even upgrade any of the hardware components. That said, if you know your way around the (ever-more complicated) web-interfaces of cloud service providers like AWS, Google Cloud or Microsoft Azure, which is no small feat, then cloud servers are relatively easy to configure and extremely simple to deploy.
Comparison: Installation & Maintenance
All servers need to be physically setup, and before they can serve a purpose like hosting a website or storing data, they also need to be digitally setup. The digital setup process of installing an operating system such as version of Linux (i.e. Ubuntu or CentOS) is the same regardless if you own a physical server or if you’re digitally connecting to one in the cloud. Where they differ is the physical installation and maintenance:
For on-premise servers, once you take ownership of the equipment you’ll need to have IT administrators on-staff, or be able to hire a third party service provider, to physically setup and configure the hardware. This includes installing and connecting the hard drives, wiring the ports and connecting the server to the network. Many medium to large companies have dedicated IT resources on staff who are well-versed in this type of work, but those that don’t as well as most small businesses will have to rely on a relatively robust nationwide network of third party installers who will come to your location and setup your equipment for you, ensure that it’s configured properly and accessible to your network.
For cloud servers, there’s no hardware to physically setup or configure as you’ll simply be renting space on an existing piece of hardware in the cloud providers data center. However, that doesn’t necessarily mean that setup is easy or straight-forward. AWS and other large providers have a grown so fast, and expanded the scope of their service offering so greatly in recent years, that their corresponding web interfaces have become incredibly complex and difficult to navigate. At the time of this publication, Amazon Web Services has over 100 distinct cloud services in which to chose from. That being said, if you know your way around their interface, the creation of VPS is a relatively quick and simple process.
Installation & Maintenance Winner: Cloud
An on-premise server will always require some amount of physical configuration and setup, and if the server bank is large or complex, will often require the presence of dedicated IT professionals to complete. The installation and maintenance of cloud servers on the other hand, is the sole responsibility of the cloud service provider and thus not the worry of the individual or business that is creating a virtual instance.
While not all-encompassing, the 2 primary security concerns as it relates to servers are access security and configuration security. Configuration security deals with how the operating system and/or system applications on your server are configured and maintained. Just like with your home computer, server operating systems and applications often need to be updated when vulnerabilities are identified, and the process for connecting to the server and making those updates (often via the command-line) can be handled equally well on both on-premise and cloud servers. Access security on the other hand, can be thought of in 2 ways:
- Physical access concerns (i.e. who can walk-up and directly access your server by plugging in and connecting to it)
- Digital access concerns (i.e. what other servers/computers can communicate with your server and potentially try and exploit it’s vulnerabilities)
Security Winner: On-premise
While both on-premise and cloud servers can be configured to limit digital access over a network, by default a cloud server will be available globally and the infrastructure used to manage the data center will almost certainly be available globally, whereas on-premise servers can be configured out of the box to only listen for and respond to requests from a particular network (like those coming from a specific office building). Furthermore, only companies with on-premise servers can 100% control who has physical access to the hardware 100% of the time.
When comparing the cost of owning a physical server with that of operating a VPS on the cloud, we need to consider 2 things:
- What are our configuration (CPU, Ram, Hard Drive) requirements?
- How long do we plan on having this server or VPS in operation?
To try and compare these disparate items in as fair a manner as possible, we will compare an entry-level Dell tower model with Microsoft Azure and a ultra-performance Dell rack server with Amazon AWS.
Entry cost comparison: Dell PowerEdge T640 Tower vs. Microsoft Azure
For the entry-level comparison, we’ve chosen the Dell PowerEdge T640 tower series as our on-premise server. This can be purchased directly from Dell for $1,459 and comes with a single-processor, 8GB RAM and 1TB of storage. To make it more comparable with the offerings from Microsoft Azure, we’ve added a second processor, which brings the total cost to $1,766.
We will compare that to a virtual instance from the rapidly growing Microsoft Azure. It’s difficult to make a perfect apples-to-apples comparison as the configurations don’t match exactly, but at the time of the article’s publication, we were able to configure a two CPUs, 8GB RAM, and 100GB of storage VPS for $185/month. Mind you that while this instance has similar computing power to the Dell tower, it only has 1/10 of the storage capacity. Take a look at the table below for the one-year ownership costs of each:
|Months||Dell T640 Tower On-premise||Microsoft Azure Cloud
Conclusion: The base Dell Tower server has 10x more storage and pays for itself in only 10 months when compared to Azure.
Performance cost comparison: Dell PowerEdge R740 Rack vs. Amazon AWS
For the performance-level comparison, we’ve chosen the Dell PowerEdge 740 Rack Server as our on-premise server. This can be purchased directly from Dell for $8,629 and we’ve configured it as follows:
- 2 Intel 4216 Xeon® Silver 16 Core (32 Core)
- 4x 16GB of RAM (64GB)
- 6x 4TB HDD (24TB)
- Purchase price: $8,629
We will compare the Dell R740 to custom-built virtual instance from Amazon Web Services (AWS) that has similar specs. As with the previous example, we can’t make a perfect comparison, but our on-demand AWS instance is configured as follows:
- Service type: EC2
- Instance: m6g.8xlarge (32 vCPUs, 128 GB RAM)
- 16 TB storage on Cold HDD
- Monthly cost: $800.59
Conclusion: The break-even time for the high-performance model is only 1 additional month than the entry-level model.
|Months||Dell PowerEdge R740On-premise||AWS m6g.8xlargeCloud
Cost Winner: On-premise
The cost of purchasing a physical, on-premise server is upfront and transparent and based entirely on simple to understand inputs like manufacturer, CPU, Ram, hard drive. While some higher-end physical servers can be quite expensive, the break-even cost of a cloud instance (when similarly configured) is almost always under 3 years and in many cases, under just 2 years.
Overall Winner & Conclusion
Without question there are many instances were a virtual private server on a cloud network makes sense. These instances are usually on the fringe, or in other words, when the server instance is very small or very, very large (think how many servers Netflix would have to own and manage to run it’s business). For most businesses that have more common use-cases however, owning a physical, on-premise server makes a lot of sense. Not only are you able to control everything from configuration and scalability to access and location, you also stand to save a lot of money over the life of the server.
If you think that a physical server might be right for your business, chat with one of our server experts today, and see if we can help you find the perfect equipment for your growing business. We offer a the latest models of Dell PowerEdge and HP Proliant servers. All are factory refurbished, which means you get the same great server for way less than it would be new, and each server we sell comes with a 24-month warranty standard at no extra charge.