Cloud and Infrastructure

What Is an Enterprise Cloud Server? 12 Criteria for Choosing One

Nanda Teknoloji Engineering Team · · 8 min read

In brief

An enterprise cloud server is a compute resource virtualized on physical hardware, scalable up and down as needed, and backed by a contract. When choosing one, look not only at vCPU and RAM but also at storage, network, backup, security, compliance, monitoring, and the support model together.

An enterprise cloud server is a virtual server created through virtualization on a cluster of physical servers, with CPU, memory, storage, and network resources allocated and scalable as needed, and with the provider taking on responsibility for the infrastructure. Instead of buying a single machine, you use a measurable resource package for your operating system and application. What makes it enterprise-grade has less to do with technical specifications than with assurances: resource guarantees, backup, access control, monitoring, compliance, and timely support. To choose correctly, you need to look not at the price list but at how your workload behaves and what the contract commits to. Below, we first cover how the model works, then the 12 criteria and common mistakes.

How does a cloud server work?

The provider installs a hypervisor on top of the physical servers. The hypervisor divides a single physical machine into multiple isolated virtual machines. Each virtual machine runs its own operating system and applications. Storage is usually delivered from a shared storage layer, and the network through virtual switches and routers. Thanks to virtualization, a virtual machine can be moved to another physical node, which provides continuity during maintenance and failures.

The core difference this structure brings is flexibility: a server can be created within minutes, its resources can be changed, and it can be shut down when not in use. In return, you give up some of the direct control over physical hardware. That is why how resources are shared and what is guaranteed become the critical questions. For its place within the service models, you can also see our article on the difference between IaaS, PaaS, and SaaS.

12 criteria to look at when choosing

  1. Resource model and guarantee. Ask whether vCPU and RAM are dedicated, or shared and possibly overcommitted. For databases that need predictable performance, dedicated resources are safer.
  2. Storage type and performance. Disk type (NVMe, SSD, HDD), IOPS, and latency values must fit the workload. Even when capacity looks sufficient, disk latency can slow down the application.
  3. Network structure. Bandwidth, private network (VLAN/VPC), static IP, load balancing, and egress traffic rules should be clear. How egress traffic is calculated is the main source of surprise bills.
  4. Availability commitment. What the service level measures under the SLA, which maintenance windows it excludes, and what happens in case of a breach must be in writing. Look at the definition, not the percentage.
  5. Backup and restore. The scope of the backup, retention period, copy to a different location, and restore testing should be determined. For details, see the backup solutions page.
  6. Security layers. Network firewall, access management, multi-factor authentication, responsibility for security updates, and DDoS/WAF options should be defined.
  7. Compliance and data residency. In which country the data is kept, who can access it, and how KVKK obligations are met in the contract should be made clear.
  8. Operating system and licensing. Supported versions, who provides the license, and how the license cost is passed on should be visible from the start.
  9. Scalability. Vertical (adding resources) and horizontal (adding servers) growth options, and API or template support for automation, are evaluated.
  10. Monitoring and visibility. CPU, memory, disk, and network metrics visible in a dashboard, threshold alerts, and log access are needed. You cannot manage what you cannot see.
  11. Support model. Support hours, response times to requests, the path to a senior engineer, and the need for Turkish-language communication should be determined. If your own team is limited, consider the managed services model.
  12. Exit strategy. Being able to take your data and virtual machine images from the provider and move them elsewhere reduces dependency risk. The process for deleting data at the end of the contract should also be in writing.

Weighing the criteria by workload

The same criterion does not carry the same weight in every workload. For a test environment, price and flexibility stand out, while for an application that processes payments, availability, security, and backup are decisive. The table below is a starting point showing which criteria carry more weight in typical workloads.

Workload Priority criteria Secondary
Corporate website Network, WAF/DDoS, monitoring Storage performance
ERP and database Resource guarantee, disk latency, backup Flexible scaling
Development/test environment Fast setup, flexibility, cost control High availability
Application processing personal data Data residency, access management, backup Network bandwidth
Application with seasonal peaks Scalability, load balancing, CDN Fixed resource guarantee

The table is not a prescription but a conversation starter. Filling in each row with one of your own real applications gives you a common yardstick when comparing offers.

Example scenario: an ERP migration decision

A mid-sized manufacturing company wants to move its ERP application, which runs on a five-year-old physical server, to a cloud server. The first offers are compared on vCPU and RAM, and the lowest-priced option stands out. When the team applies the criteria list, it notices three things: the database disk is on the shared tier and no latency value is stated, the backup is kept only in the same location, and support is provided only during business hours. A problem at month-end close could have turned the cheap-looking offer into an expensive risk. The team chooses the offer that includes disk performance and an off-site backup, and sizes the migration using the previous month’s load data.

This example shows the function of the criteria list: it makes offers comparable along the same axis and makes the missing commitments visible.

A method to follow when comparing offers

The criteria list only works if offers are compared on the same ground. A short evaluation process helps with that.

  1. Define the workload. Which applications will be migrated, and what are the average and peak CPU, memory, disk, and network usage? Without this data, each offer is prepared with a different assumption.
  2. Send the same question form to all providers. Turn the 12 criteria into questions and get the answers in writing. Verbal explanations are not remembered later.
  3. Separate commitment from feature. “We support it” is a feature; “we respond within this time, and in case of a breach we do this” is a commitment. Only commitments go into the contract.
  4. Run a small pilot. Run a copy of your real application and measure the disk, network, and support experience. A pilot tells you more than a presentation.
  5. Work out the total cost. Add licensing, backup, additional IPs, security services, egress traffic, and operating effort to the server fee.

Internal stakeholders also need to be involved early in this process. The application owner clarifies performance expectations, the security team access and log requirements, the legal team the contract and data location clauses, and the finance team the payment model. A requirement that emerges later weakens your negotiating power after selection.

Shared responsibility: who manages what?

The most commonly misunderstood topic with cloud servers is the responsibility boundary. In general, the provider manages the physical infrastructure, the virtualization layer, and the infrastructure network; the operating system, application, data, and access permissions are mostly the customer’s responsibility. However, this boundary can shift depending on the contract. For example, in a managed services model, operating system patching, monitoring, and backup oversight can also be handed over to the provider.

Before signing the contract, get clear answers to these three questions: Who applies operating system security patches, and how often? In the event of a failure or security incident, who makes the first response, and how soon do they notify you? Who monitors whether the backup succeeded, and who tests the restore? If the answers to these three questions are unclear, your workload has no real operations owner.

Common mistakes

  • Looking only at vCPU and RAM. Disk, network, and support are often the real bottleneck.
  • Sizing without measuring current usage. Resources chosen without average and peak usage data from the physical server end up either excessive or insufficient.
  • Assuming backup is included. Backup scope and restore time are not guaranteed unless they are in writing.
  • Not accounting for exit cost. Egress traffic and data transfer fees can change the total cost.
  • Leaving the security responsibility unclear. If it is not determined who owns operating system patching, unpatched systems remain.
  • Not having an exit plan. If portability is not considered from the start, the cost of dependency emerges later.

Decision summary

It helps to approach cloud server selection in three steps. First, measure the peak and average resource usage of the current workload. Then weight the 12 criteria by the criticality of the workload and score each offer in the same table. Finally, read the commitments in the contract, especially the SLA definition, the backup scope, and the exit terms. For some workloads, a cloud server may not be the right answer; we cover this in our comparison of physical server or cloud server and, for hybrid scenarios, in the hybrid cloud architecture article.

Moving forward with Nanda Teknoloji

Nanda Teknoloji carries out the discovery, design, migration, and operations steps in projects like this: measuring the current load, sizing, security and backup design, migration, and monitoring afterward. You can review the cloud server options and their scope on the cloud server solution page, and talk to our team for an assessment specific to your workload.

Frequently asked questions

What is the difference between a cloud server and a VPS?

Both are based on virtualization; the difference shows up in resource guarantees, scaling flexibility, management of the network and storage infrastructure, and contractual assurances. An enterprise cloud server expects more detailed and measurable commitments in these areas.

Does the application need to be rewritten to move to a cloud server?

Most of the time, no. With a virtual-machine-based migration, the application can be moved as is; re-architecture is considered only if a scaling or cost goal requires it.

Is backup included with a cloud server?

Do not assume it is included by default. Backup scope, retention period, and restore testing should be clarified separately in the contract.

Does increasing resources later require downtime?

This depends on the operating system and the virtualization layer. Some changes can be made while running, while others require a restart; ask the provider before the migration.

Related solutions

Related posts

Cloud and Infrastructure

Hybrid cloud architecture: When physical, when cloud?

What is hybrid cloud, and which workload belongs on a physical server or in the cloud? A decision framework covering network, data, security, and operations.

· 8 min read

Let's plan your critical systems together

Tell us what you need; our team will work out the right architecture and migration path with you.