Indeed, for decades, putting workloads in the cloud was seen as the natural choice for growing businesses. No need to invest in hardware in advance, getting a new server in minutes, giving small IT departments access to technology that could only be afforded by larger organizations – the benefits were obvious. The same logic applies to many workloads today. However, an increasing number of businesses are taking a closer look at their infrastructures after a couple of years of cloud investments, when the initially small bills become hard to justify.
This process doesn’t always end up with yet another optimization of cloud solutions. Certain UK companies are shifting stable databases, web applications, internal platforms, and other predictable workloads to dedicated servers. Companies like BlueServers operate in the hosting segment that has gained relevance lately as more businesses start questioning whether all of their workloads require cloud elasticity. The problem here is not whether cloud or dedicated infrastructure is better. Instead, the issue is whether specific applications would benefit from paying for consumed resources or a fixed amount of hardware per month.
The cloud bill may prove challenging to predict
Public cloud bills are not simply bills on virtual servers. Compute units, storage, snapshots, databases, bandwidth, load balancing, monitoring, and data transfers are billed independently. One service may not even add up much. Many combined together in a stack can be hard to predict.
It is important to small businesses since those have small infrastructure teams in place. It does not mean that five engineers in a company can afford themselves the luxury of analyzing the reserved compute resources, unused volumes, and network costs weekly.
This problem exists outside of SMEs as well. Flexera’s “State of the Cloud 2026” study revealed that 85 per cent of its participants identified managing costs as their number one concern when it comes to working in the clouds. That same study suggested 29 per cent wastage in IaaS and PaaS spending.
Think about a popular website selling products online. In November and December, the flow of visitors may grow sharply, making cloud scaling essential during these periods. The database, accounting software, and internal reporting system of this company may operate at a more or less uniform pace throughout the year. It may be quite expensive to keep these steady loads in a cloud environment.
A dedicated server will provide a different scenario. The company pays a fixed fee for a certain number of CPUs, a certain quantity of RAM, certain storage resources, and certain bandwidth. This way, financial planners of the company will have no trouble predicting its costs for one year ahead.
Consistent workload favors physical hardware
The ability to scale up cloud infrastructure addresses an important issue because resources can be scaled up easily. This feature is very valuable in case of unpredictable demand. It matters less if the application has been running for three years and its resource needs are known.
Physical hardware eliminates the need for a virtual machine in the basic computing model. The company will be aware of what processors, disks and memory are used by the server. It can come in handy for workloads that constantly utilize large amounts of CPU, RAM or disk input/output operations.
Examples include:
- large databases with sustained query loads
- busy e-commerce back ends
- video processing and media workloads
- data analysis jobs that run for hours at a time
- business applications with predictable daily traffic
Performance does not necessarily get quicker just because the server is dedicated. It depends on the type of storage, network, software and even the age of the hardware. The important thing is consistency. The firm will be able to build the machine based on the actual needs and know what resources are there.
This will help to make the capacity planning tangible. For example, if a database uses 18 GB of memory in the peak period, an administrator will have a choice between a 32 or 64 GB machine, instead of changing the settings of the cloud machine all the time.
The place of data storage requires more than a postcode
Data sovereignty is another reason for UK firms to look at their infrastructure more carefully, especially in case of customer records, personnel data or payment systems.
The fact that there is a server in London does not in itself answer all questions under UK GDPR. The Information Commissioner’s Office draws an important distinction between the location of servers and the organisation which is receiving or accessing personal information.
According to the existing ICO guidance, the storage of information on servers located outside of the UK does not always mean that there is a restricted transfer involved. Businesses must determine which legal entity is providing the service, in which country this organisation is registered, whether foreign organisations or subprocessors have access to the information and what type of transfer applies.
This information is easily overlooked while selecting infrastructure on the price page.
Thus, a UK business wishing to use dedicated hosting services must know where exactly the physical server is located and who is operating it. The business must also know who has access to backup systems, monitoring and support.
For any business processing large amounts of personal information, procurement of infrastructure should include the person responsible for the protection of personal data rather than be considered only as an IT task.
What to check before signing the contract
Shifting workloads from a public cloud doesn’t mean the need for infrastructure management vanishes. Someone still needs to take care of operating systems, perform security patches, keep track of storage, make backups, and deal with hardware failure.
This level of responsibility greatly depends on the hosting arrangement. For example, a cheap dedicated server can be just a server in a rack with electricity and connectivity provided. Managed hosting can include monitoring, operating system management, replacement hardware, backups, and support.
Here are some things a business should clarify before signing the contract:
- the exact CPU, RAM and storage supplied
- whether drives use RAID and what happens after a disk failure
- replacement times for failed hardware
- included bandwidth and any traffic limits
- backup location and retention period
- DDoS protection and network security
- support availability outside normal working hours
- data-centre location and provider legal entity
- access available to subprocessors or overseas support staff
- contract length, cancellation terms and migration assistance
The network’s capability should also be paid special attention to. A server with top-class processors and NVMe drives will deliver poor results if the link becomes a bottleneck.
Questions about backup systems should be specific. The fact that a provider offers some backups is not enough. Companies need to find out how often backups take place and where the copies are located.
Cloud repatriation does not mean undoing everything businesses have been taught about cloud computing. Those companies planning to go for dedicated infrastructure do not plan to bring all their applications back from the cloud. They are getting more selective.
It can make a lot of sense to use a short-term development environment in the public cloud. A service with sudden spikes in traffic needs elasticity. SaaS solutions eliminate whole layers of infrastructure effort. At the same time, a mature product with stable load, known performance characteristics and terabytes of predictable data can be easily operated in the dedicated infrastructure.
This is a distinction that becomes more critical as infrastructures grow up. The pertinent query in the case of an UK-based company will not be only in where its operations should reside. It has to do with the real requirements for each workload and whether this flexibility is being used and paid for.
