Telco Cloud

ACG Research: Save 40% on Total Cost of Ownership when Building a Virtualized, Automated Network

Communications Service Providers can save 40% of their costs by using a virtualized, horizontal network environment with automation as compared to a non-virtualized, “vertical” environment  

“The adoption and success of NFV,” writes the Institute of Electrical and Electronics Engineers (IEEE) in a 2022 article titled VNF Software Cost Modeling Based on Telecom Network, “are contingent on managing software costs and resources. It involves multiple conflicting objectives, such as energy, licenses, resource migration and cost.”1 NFV, or Network Function Virtualization as the article elaborates, “integrates cloud and virtualization technologies to rapidly develop new network services while enhancing flexibility, scalability, and automation”2 and constitutes a critical component of successful 5G network modernization and improved service time-to-market. Understanding the telecom industry’s imperative to provide 5G-based services and the importance of these virtualization and cloud-native technologies in enabling them, VMware partnered with ACG Research—a technology consultancy—to explore how reducing the Total Cost of Ownership (TCO) of one’s network can set up Communication Service Providers (CSPs) for ongoing financial and operational success.  

The results of our study are clear: with a virtualized, “horizontal” network environment complemented with the right automation technology, a CSP can save roughly 40% of their costs in a five-year period when compared to a non-virtualized, “vertical” environment. It is necessary, then, that we elaborate on the benefits of those two financial drivers—“horizontal” architecture and automation technology—as those components are necessary for any CSP seeking substantial cost savings, operational efficiencies and competitive differentiation. Figure 1 below offers an overview of the deployment scenarios we modeled: 

Figure 1 

In the following, we elaborate on the two essential drivers of total cost of ownership—capital expenses and operating expenses—when a CSP is building its network. The capital expense discussion focuses on a critical decision for CSPs: to deploy one’s network “horizontally” or “vertically.” We argue that a “horizontal” choice reduces capital expense obligations by requiring fewer servers and smaller data centers to power one’s network operations. Operating expenses, on the other hand, focus on the benefits of an effective automation platform that streamlines manual tasks and allows skilled engineers at a CSP to focus on developing innovative network services instead of remedying internal network operations. 

“Horizontal” vs. “Vertical” Deployments: Drivers of Capital Expenses 

When constructing a network, CSPs often face a crucial decision: create a network where core data centers, RAN and edge locations are deployed with a series of vertically-integrated servers, or a network where these sites are deployed with servers that can dynamically share and pool resources. A visualization of the architectural differences is offered in Figure 2 below: 

Figure 2 

Of course, each deployment choice has its own benefits. A vertical network deployment allows the CSP to install dedicated, vertically-integrated stacks into specific geographies or network domains to satisfy an immediate purpose. If, as a CSP, you’re seeking a readymade stack of hardware and software dedicated for one purpose, then a vertical approach makes sense—a quick-fix to inexpensively bandage a network requirement that, given the vertically-integrated nature of a NEP stack, requires few network function integration challenges.  

Like many decisions in life, however, the tradeoff between short-term satisfaction and long-term pain (e.g., “if I eat a Big Mac for dinner tonight it will taste better than a salad but won’t do me any favors in the long run), this vertical approach poses major shortcomings. Vertical stacks, like sports cars or fighter jets, serve single purposes that assign resources to applications upon request. Simply put, if you require an additional function for your network, it necessitates a new server. Let’s use Microsoft Word as an illustration.  

When you open Microsoft Word on your personal computer, the application requests a specific level of computing resources from your operating system. If the resources are available, the application will start. If not, you must wait and waste valuable time until the resources free up. With this resource assignment process, resources are depleted as more applications are started. Without sharing (i.e., consolidating) physical resources, as seen in horizontal environments, you require more resources to run applications which necessitate more servers, electricity, and maintenance—driving up operating and capital costs.  

Kubernetes, a platform that runs on a Linux operating system commonly used in network deployments, functions in the same way as our Microsoft Word example above. When a CSP executes a workload to support an application, resources are assigned to Kubernetes containers. If, however, assigned resources are unavailable, the containers will not start in the Kubernetes cluster and the network function or application will fail—as such, as networks expand and their requirements burgeon, it’s financially and operationally infeasible to continuously install new stacks to serve discrete application requirements. It comes as no surprise, then, that according to IEEE’s study, network TCO reduces when “resources share, or consolidation has taken place”3 as found in virtualized, horizontal environments.      

Horizontal stacks, as depicted on the left-hand side of Figure 2, utilize an important differentiating technological aspect not found in vertical stacks—a hypervisor. Hypervisors, the SDDC/ESXi layer on the left-hand side of Figure 2, virtualize underlying server infrastructure so that the Kubernetes layer (for container-based network functions, CNFs) or Virtualized Infrastructure Manager (VIM) (for virtualized network functions, VNFs) share physical resources so that fewer servers can power multiple functions and applications. With a hypervisor, in other words, resources are shared, which means that applications have access to resources, but the hypervisor can manage and reallocate resources if other work is required. Hypervisors importantly manage resource requests to virtually assigned resources or hardware without compromising network performance and enable resource sharing and workload consolidation—driving high server utilization, requiring fewer servers to execute critical network tasks and reducing costs associated with server acquisition, installation and maintenance. IEEE, no wonder, argues that this “sharing minimizes the TCO with optimum resource utilization”4 of servers in one’s network. In fact, our research with ACG found that an automated horizontal deployment saves 18% in capital expensesover five years and reduces server requirements by 36% when compared to a vertical approach—presenting substantial financial savings and, with a reduced data center footprint, creates a greener network as well when deploying horizontally.   

To Automate or Not: Drivers of Operating Expenses 

In a joint study published in February 2023, market research firm, Analysys Mason, and Google concluded that network automation is “essential to realising key benefits of cloud-native networks, such as low total cost of ownership […]”5 In fact, the study which involved interviewing scores of CSP executives, department heads and managers found that “[a]utomation is the top strategic initiative for CSPs […]”6 and “CSPs need standard approaches to automation to combat siloed technologies and different vendor approaches.”7 The argument for network automation, in theory, feels obvious: effectively managing a network requires successfully executing dozens of operational tasks—why spend valuable time and resources addressing these tasks manually when a centralized automation and management platform can tackle the same tasks with greater speed and precision? The reality is, however, more complicated: automation platforms carry upfront costs and potentially face integration or interoperability issues with legacy systems. With these complications in mind, our research led to understanding how well-constructed automation platforms can reduce one’s total cost of ownership. 

When modeling the benefits of an effective automation platform, we wanted to include a platform capable of the following: 

  • Provides holistic lifecycle management for Days 0, 1 & 2 operations as well as ongoing application and network service management and automation​ 
  • Pre-built integrations with complementary products (e.g., Kubernetes orchestration, infrastructure virtualization)​ 
  • Dynamically customizes underlying virtual infrastructure to support changing network function requirements​ 
  • Supports any network function from any vendor​ 
  • Centralized multi-cloud, multi-domain management with a single pane of glass​
  • Aligns with industry standards (e.g., ETSI) 

As such, a robust automation platform reduces manual operational errors, fully automates workflows, improves maintenance efficiency, increases action success rates and allows engineers to focus instead on strategic, revenue-generating initiatives—all driving down operating expenses. In fact, according to our research, utilizing an automation platform like the one described above reduces total cost of ownership by 37% and operating expenses by 66%, compared to not using an automation platform to build one’s network.8    

Wrapping it Up

A combination of ACG’s market data and insight as well as VMware’s own customer experience led to a fruitful research collaboration that allowed us to accurately depict the financial value (i.e., cost savings) of a horizontal, virtualized architectural choice over a vertical, siloed one as well as the incremental financial benefit of using an automation platform to deploy one’s network. 

When evaluating a vertical approach against a horizontal one with automation, our research produced the following results: 

  • 40% total cost of ownership savings over five years  
  • 36% reduction in servers required for your network, which lower capital costs and drive down carbon emissions at data centers by nearly 30,000 metric tons of carbon dioxide over five years 
  • 62% savings in operating expenses and 18% savings in capital costs over five years 
Figure 3 

We encourage you to read the corresponding white paper from ACG here. The white paper provides additional details on our project goals, methodology, and the main drivers of operating and capital savings. For additional information on the horizontal automation platform we modeled, learn more here. 


Sources:
 1 – G. Bista, E. Caron and A.L. Vion, “VNF Software Cost Modeling Based on Telecom Network,” 2022 Ninth International Conference on Software Defined Systems (SDS), Paris, France, 2022, p. 1. 

2 – G. Bista, E. Caron and A.L. Vion, “VNF Software Cost Modeling Based on Telecom Network,” 2022 Ninth International Conference on Software Defined Systems (SDS), Paris, France, 2022, p. 1. 

3 – G. Bista, E. Caron and A.L. Vion, “VNF Software Cost Modeling Based on Telecom Network,” 2022 Ninth International Conference on Software Defined Systems (SDS), Paris, France, 2022, p. 8. 

4 – G. Bista, E. Caron and A.L. Vion, “VNF Software Cost Modeling Based on Telecom Network,” 2022 Ninth International Conference on Software Defined Systems (SDS), Paris, France, 2022, p. 1.

5 – Google Cloud and Analysys Mason, “Cloud-Native Automation: The Transformation of CSP Networks,” February 2023:  p. 1.

6 – Google Cloud and Analysys Mason, “Cloud-Native Automation: The Transformation of CSP Networks,” February 2023: p. 2, emphasis added.

7 – Google Cloud and Analysys Mason, “Cloud-Native Automation: The Transformation of CSP Networks,” February 2023: p. 2.

8 – Note: the vertical scenario we modeled also includes an automation and orchestration solution. See the link to the white paper in the conclusion to learn more about each scenario’s specific inputs and modeling methodology.