Cost efficiency in software development through cloud computing

Currently, there is a rapid development and widespread adoption of cloud computing technologies. According to annual surveys conducted among IT directors regarding technological investments, cloud computing has risen from the sixteenth to the second position [1, p. 2139-2155]. Back in 2011, experts suggested that cloud computing, along with mobile technologies and big data processing, constitutes the "third platform" in the IT industry, following the eras of mainframes and personal computers (PCs) [2]. Companies failing to recognize the significance of cloud technologies may soon face market exit, similar to businesses in the 1980s that did not transition from mainframes to PCs [2].

Today, IT leaders are primarily concerned with assessing the feasibility of adopting cloud platforms and evaluating the economic benefits and risks associated with their implementation. This involves addressing the following key tasks:

• selecting the appropriate type of cloud provider services;
• estimating short-term and long-term costs of implementing cloud computing;
• calculating the profitability of cloud adoption in both short and long terms;
• analyzing the risks associated with cloud adoption and those stemming from retaining legacy IT strategies.

Such evaluations must consider the company's business specifics and market tenure. To achieve these goals, it is essential to have a clear understanding of cloud computing and comprehensive information about cloud providers and their services.

Cloud computing is a distributed data processing technology where computing resources and capacities are provided to users as an internet service. A cloud service relies on a client-server model, allowing users to access server resources organized as follows:

• the server group appears as a single virtual server to the user;
• users can flexibly adjust resource consumption based on their needs;
• scalability is ensured to handle increased loads;
• risk of server failure is minimized through automatic resource reallocation in case of a malfunction.

Figure illustrates the classification of cloud computing based on deployment models (private, public, and hybrid) and service models (IaaS, PaaS, and SaaS). This classification highlights the versatility and applicability of cloud solutions across diverse organizational needs.

Fig. Classification of cloud computing based on deployment and service models

In global practice, the development of cloud computing has become a large-scale trend and is a central focus in the development plans of the United States and Europe. However, according to Sharma et al., cloud computing is merely a marketing ploy and a new term for virtualization and on-demand services [4].

However, cloud computing is more than just the rebranding of old technologies. Until recently, virtualization software – particularly for x86 architectures – was almost monopolized by VMWare, and server hardware was prohibitively expensive [5]. This restricted the use of such technologies to large enterprises. The emergence of free and open-source virtualization software (e.g., Freeware versions for x86 and x86-64 architectures by leaders such as VMWare and QEMU), free APIs for developing distributed applications (e.g., PHP, JSP, ASP.NET), significant cost reductions in blade servers, and the mass adoption of 64-bit architectures have made cloud technologies far more accessible. This democratization of previously expensive software and hardware for creating powerful data centers has led to a proliferation of commercial data center offerings. At the time, there were no converged solutions that combined cutting-edge technologies into commercially viable products for companies and private users. New cloud vendors like Amazon, Google, and Microsoft Azure successfully addressed this gap with their innovative offerings.

Cloud infrastructure can be categorized into the following types [1, p. 2139-2155; 3; 6]:

1. Private Cloud: Essentially a dedicated cloud data center owned by an organization.
2. Public Cloud: A collection of cloud services provided by a third-party organization (cloud provider).
3. Hybrid Cloud: A combination of private and public cloud solutions.

Services provided by cloud providers include:

• SaaS (Software as a Service);
• PaaS (Platform as a Service);
• IaaS (Infrastructure as a Service);
• MaaS (Monitoring as a Service) – a new direction [7, p. 123-136];
• CaaS (Communication as a Service) – another emerging model [7, p. 123-136].

The advantages of cloud computing make it highly attractive for cloud service providers. Key benefits include:

• Reduced costs for data center setup and maintenance.
• Faster time-to-market without the delays of building IT infrastructure.
• Accessibility of high-performance applications for small and medium-sized businesses.
• Unlimited scalability and flexibility.
• Enhanced reliability and operational continuity.
• Increased mobility for employees globally.
• A shift in IT departments' focus toward innovation and development rather than routine data center maintenance.
• "Green" data centers, achieved by minimizing idle capacities.

Public cloud providers deliver services across multiple regions and zones within each region. This geographic distribution enhances the responsiveness of applications by allowing businesses to deploy resources closer to their customer base. Furthermore, by deploying redundant systems across zones within a region, applications can ensure continued operation even if a catastrophic event disrupts a single zone. Achieving similar functionality in on-premises data centers (ODCs) would require building and maintaining multiple ODCs, significantly increasing costs, potentially by orders of magnitude. This underscores a key economic and operational advantage of public cloud solutions.

The economic advantages of cloud computing are undeniable, particularly for startups or companies undergoing significant expansion. These organizations can avoid expenses associated with building and maintaining data centers, disposing of outdated equipment, acquiring new infrastructure, adapting applications to legacy systems, and retraining employees.

Public cloud providers offer substantial cost savings for businesses that reserve long-term capacity. By committing to a specified level of resource usage over extended periods, organizations can secure lower pricing tiers. This approach not only reduces operational costs but also enhances budget predictability for businesses. Such long-term reservation options further differentiate public cloud solutions from traditional ODC models, where similar cost optimizations are typically unavailable.

Data center setup and maintenance are typically the largest cost items for a company, often exceeding 50% of total expenses [3]. However, not all companies are startups or rapidly growing enterprises. For established businesses, adopting cloud computing involves both economic benefits and significant challenges. On one hand, they must invest heavily to restructure their operations and accept risks that are not yet fully understood. On the other hand, they must assess whether forgoing cloud computing will compromise their competitiveness in the future. Addressing this question requires solving the previously outlined tasks.

The Table 1 below outlines potential scenarios for adopting cloud technologies:

• Developing an application from scratch using traditional models.
• Developing an application directly with cloud technologies.
• Migrating an existing application entirely to the cloud.
• Continuing to use a non-cloud-based application without modification.

Each scenario involves both one-time and recurring expenses. One-time expenses occur during the development and implementation phases, while recurring expenses persist as long as the application remains operational (table).

Table

Costs under different cloud computing scenarios [3]

Based on the data presented in Table 1, adopting cloud technologies offers several advantages over traditional development and application maintenance models. While traditional models incur high one-time costs due to equipment procurement and application development, cloud-based approaches significantly reduce these initial investments. However, cloud migration and development involve moderate one-time costs for application adaptation and training.

Recurring costs also differ significantly between the scenarios. Traditional models entail high expenditures for premises and infrastructure maintenance, alongside salaries for in-house technical staff. In contrast, cloud-based models shift a significant portion of recurring expenses to service rentals and technical support, which are generally scalable and predictable.

Notably, cloud migration combines the benefits of reduced infrastructure costs with scalable service expenses, making it a cost-effective solution for organizations seeking to modernize legacy systems. However, decision-makers must account for medium recurring costs, such as service rental and support fees, that could accumulate over time.

Adopting cloud technologies eliminates costs related to purchasing and maintaining proprietary equipment and reduces the need for salaries of staff primarily responsible for hardware operations rather than application management.

For existing systems, transitioning to cloud technologies removes recurring costs for maintaining proprietary equipment. However, this shift incurs expenses for reworking applications and decommissioning in-house data centers. Depending on the specific tasks and conditions of each organization, any of the aforementioned scenarios could be the most economically viable. Nonetheless, cloud technologies are particularly attractive as they free up funds that would otherwise be allocated to non-core IT tasks.

Profitability calculation for cloud computing adoption

In [8, p. 9574-9603], a formula is proposed for calculating the economic benefits of adopting cloud technologies:

, (1)

Where:

• : Hours of cloud usage.
• : Hours of own data center (ODC) usage.
• Tr: Revenue generated.
• : Cost per hour of cloud usage.
• : Cost per hour of using own ODC.
• U: Average utilization (load factor) of the own data center.

The formula assumes that:

• A private cloud is equivalent to the company's own data center.
• An external cloud refers to services provided by third-party cloud providers.

If the inequality holds true, it indicates that the company can achieve equal or higher profits by adopting cloud services instead of using its own data center (ODC).

The comparison depends on:

• The cost per hour of cloud usage versus own data center usage,
• The utilization efficiency of the company's data center (𝑈),
• The expected revenue (𝑇𝑟).

Ultimately, this formula helps decision-makers evaluate whether transitioning to cloud solutions is economically advantageous compared to maintaining their own data center infrastructure.

The left side of Equation (1) represents the revenue a company can generate by using cloud computing, calculated based on a specified number of machine hours. The right side of Equation (1) reflects the revenue the company can obtain from operating its own data center (DC).

The main distinction lies in the consideration of the average utilization factor for the DC in Equation (1). For example, if the DC operates at only 10% capacity, the cost per hour of DC operation is multiplied by 10, which significantly reduces the overall revenue in this scenario.

An economically ideal situation is when the average utilization of the DC approaches [8, p. 9574-9603]. However, Equation (1) does not account for the following factors:

• The comparison period for revenues (the left and right sides of the inequality) must be identical.
• A 100% average utilization is practically unachievable, as it would indicate catastrophic overload. The optimal utilization typically ranges between 60% and 70%.

The cost of one hour of DC operation already includes overhead costs incurred from above-average utilization. Idle equipment remains connected, software is installed on it, and it requires maintenance while consuming electricity. Additionally, staff is hired based on total capacity rather than actual utilization. Thus, when dividing the costs per hour of DC operation by average utilization, Equation (1) unjustifiably inflates the DC costs.

To address these factors, the cost per hour of DC operation can be replaced with the average expenses for utilized DC capacities, i.e., excluding the costs of maintaining surplus capacity. However, the issue is that the cost of utilized DC capacity is not an obvious parameter for the company and requires computation, while explicit data is only available for total DC expenses.

Based on the above considerations, the following revised inequality is proposed to assess the economic benefits of cloud computing.

Based on the considerations outlined above, the following inequality is proposed to assess the costeffectiveness of cloud computing:

, (2)

Where:

• represents the revenue generated using cloud services over the period ;
• represents the revenue generated using the company's own data center over the period ;
• represents the expenses incurred for cloud services over the period ;
• represents the comprehensive expenses of the data center over the period , including the average cost of DC equipment (calculated as the initial equipment cost divided by the total usage cycle and multiplied by the relevant period), as well as the costs of maintaining the DC during the same period, which account for employee salaries, software acquisition, and utility bills.

In Equation (2), as in Equation (1), a data center owned by the company is not classified as cloud computing, even if it is highly capable and exhibits cloud-like characteristics (it is designated as a DC rather than a private "cloud"). If possible, it is preferable to perform the calculation over the same period for both sides of the inequality, as this yields more accurate results. This approach is particularly straightforward when part of the business operates using the company's own DC, while another part relies on cloud service providers. For long-term benefit assessments, a longer period should be considered, while for short-term evaluations, a shorter period is more appropriate. It is advisable to use an average annual demand period, which is especially relevant for "seasonal" businesses, such as one year.

The formula (2) can be modified as follows:

, (3)

As previously noted, the average utilization rate, used as a divisor to account for the factors mentioned in formula (2), should not be included in formula (3). This is because the value of already reflects an inflated cost over the required period (including surplus capacities, employee salaries, and maintenance expenses for unused capacities during this period). However, the equation can be expressed in terms of average utilization for clarity, particularly benefiting companies with seasonal business models.

If an optimal utilization rate of 60% for the DC is assumed, then:

, (4)

Where:

• is the average utilization rate over the period, %, with ;
• is the cost of required DC capacities for the period;
• is the cost of the actual utilized DC capacities for the period.

Equality (4) is approximate because the costs of utilized and unused DC capacities are generally not identical (for instance, the cost of unused capacities may be slightly lower due to reduced power consumption and minimal equipment depreciation).

Accordingly, the surplus expenses for maintaining the DC over the period can be calculated as:

,

This formulation highlights the additional costs incurred from unused capacities, providing insight into potential inefficiencies in DC operation.

A cloud service provider can offer lower costs by minimizing surplus expenses due to the high average utilization rate of their infrastructure, which results from the scalability of their data centers and the large number of users. As the average utilization rate approaches the optimal level (60%), surplus expenses become zero.

Let us modify Equation (3) to account for the average utilization and surplus expenses, demonstrating the inverse proportionality between utilization rate and expenses, and hence the cost-effectiveness of cloud services under conditions of uneven utilization:

,

For .

If , system failures may occur, leading to the following scenarios:

1. A reduction in revenue proportional to a loss coefficient :

,

Where is the revenue during the critical period when the DC is overloaded above 60%, and is the revenue during normal operation.

2. A complete loss of revenue, where .

3. Significant financial losses, where .

By incorporating the average utilization rate and surplus expenses, Equation (3) highlights the inverse relationship between utilization rate and expenses, demonstrating the advantage of cloud computing for businesses with uneven workloads. The equation remains:

, (5)

For .

In cases where , the risk of system failure increases, leading to potential revenue losses or operational disruptions.

The higher the utilization rate, the lower the costs of maintaining a data center (DC). According to Equation (5), public cloud services are more cost-effective at lower average utilization rates of a DC. Furthermore, opting for cloud computing is also advisable when average and peak utilization rates cannot be accurately predicted. Thus, the conclusions derived from Equation (1) also apply to Equation (5). The advantages of cloud computing, similar to those in Equation (1), are evident:

• For startups, the cost per hour of operating their own DC is extremely high due to significant initial expenses for equipment procurement and hiring personnel.
• For seasonal businesses and those with highly variable DC loads (e.g., tour operators, online stores).
• For businesses with low or poorly predictable hardware utilization. If the actual load is significantly lower than expected, the company will avoid overpaying for idle capacities. Conversely, if the load exceeds expectations, adding the required resources in a cloud environment is far easier than purchasing additional equipment for a DC.
• For well-established companies, the benefits are also significant. According to IT analysts, the average DC utilization rate is approximately 18%, and for x86 server architectures, this figure drops to 12% [1, p. 2139-2155]. For such companies, transitioning to cloud computing is recommended gradually, starting with smaller projects and moving on to larger implementations. This approach minimizes risks and "shock situations" while allowing companies to evaluate the actual cost-effectiveness of cloud computing using Equation (6).

It is worth noting that Equation (5), unlike Equation (1), accounts for the optimal utilization rate of a DC. It recognizes that a utilization rate above 60% is critical and may lead to DC failures, in contrast to Equation (1), which considers 100% utilization as not only non-critical but optimal.

Autoscaling, a feature provided by public cloud platforms, enables applications to dynamically adjust their resource usage in response to traffic or workload fluctuations. During periods of low load, resources scale down, significantly reducing costs. Conversely, when traffic spikes, resources scale up incrementally to meet demand without overprovisioning. This capability ensures optimal resource utilization and cost efficiency, contrasting sharply with traditional ODC setups that must provision resources for peak load conditions, resulting in underutilization and higher costs during low-demand periods.

Realities of cloud computing adoption

The economic benefits of cloud computing are evident from the formulas for assessing cost-effectiveness. However, recent research conducted by Insight Express, surveying over 1,300 IT leaders in Australia, Brazil, the UK, Germany, India, Spain, Canada, China, Mexico, Russia, the US, France, and Japan, paints a different picture [9].

The shift to private or public cloud technologies is so daunting for many IT leaders that over a third of respondents (39%) admitted they are not ready to address this challenge. Nevertheless, nearly three-quarters (73%) expressed confidence in their capabilities, stating they have the resources necessary to implement public or private cloud technologies.

These findings contradict earlier forecasts published in document [10], which predicted that by 2024, over 50% of DC computations would be cloud-based and that annual cloud traffic would grow more than twelvefold by 2025 (to 1.6 zettabytes).

This raises the question: why are companies hesitant to adopt cloud computing? The most significant obstacle to adopting cloud computing, especially public cloud solutions, is security. As cloud services are increasingly used for strategic and critical business applications, security has become the top priority. Key aspects of cloud computing requiring improvement include:

• Monitoring: 17%,
• Availability: 17%,
• Management: 27%,
• Security: 39%.

The vast majority of clients, when learning about cloud technologies, state that they are more inclined to create a private DC on their premises due to a lack of familiarity with public cloud security and solutions to address these concerns.

Some analysts believe that ensuring cloud computing security will become a primary focus for IT vendors in the near future [11].

When using public clouds, the responsibility for security falls entirely on the provider. Moreover, businesses relying heavily on public clouds become dependent on the provider's behavior and success. If the provider exits the market for any reason, companies face the risk of business disruption or significant financial and time costs associated with transitioning to another cloud platform.

Unfortunately, there are currently no clear international standards or legal regulations governing cloud computing, as these are still under development. Cloud providers often fail to provide detailed information about security measures. Despite the use of secure protocols such as SSL and SSH for data transmission, issues related to authentication remain insufficiently addressed. Trust in cloud operators is also a concern, as there is a risk of data misuse by unscrupulous providers.

While secure protocols like SSL (Secure Sockets Layer) and SSH (Secure Shell) are used to protect data during transmission, the following specific concerns are highlighted:

1. Weak or inadequate authentication: The systems or users accessing cloud services may not have robust mechanisms to verify their identities, leading to potential vulnerabilities.
2. Risk of unauthorized access: Without strong authentication processes, malicious actors could exploit weak entry points to gain unauthorized access to sensitive data or cloud systems.
3. Reliance on providers: Businesses relying on public cloud providers might lack transparency or guarantees about how authentication is enforced and whether it meets the necessary security standards.
4. Data integrity and confidentiality: Poor authentication can undermine the integrity and confidentiality of data, even if the transmission protocols are secure.

Conclusions

The economic attractiveness of public cloud computing is evident when compared to traditional models of hosting computing resources on-premises. This is especially true for startups and businesses with uneven or unpredictable workloads, where forecasting maximum possible loads during business growth, downsizing, or transitioning to new activities is challenging. In scenarios of unexpectedly high loads, cloud services offer scalability to prevent significant losses or even business failure.

However, despite the clear advantages offered by cloud providers, businesses are still hesitant to fully transition to cloud platforms. The unresolved issue of cloud security remains the primary challenge to be addressed in the near future.