By Ajay Gavagal, Comcast; Mehul Patel, Comcast; Sinan Onder, Comcast
IP networks are designed to support a wide range of services, including video streaming, high-speed data transmission, voice communication, and more. However, any advancements or updates in either software or hardware necessitate the costly and time-consuming process of upgrading vendor-specific appliances deployed in these networks, posing significant challenges for network operators. In recent years, network operators have critically reassessed the conventional approach of relying on vendor-specific appliances that tightly integrate hardware and software to deploy and operate their networks. As a result, the concept of virtualizing network functions has gained widespread acceptance among network operators. Virtualization allows specific functions of the network to be decoupled from proprietary hardware and instead implemented as software-based on virtual instances. These virtual instances utilize commercial off-the-shelf (COTS) hardware and open new possibilities for network operators. The virtual instances could either be in a private or a public cloud (consider on-demand scaling, disaster recovery scenarios). Management and operation of networks requires a whole host of software applications that can be run in a private or a public cloud as well. Network operators have gained several advantages while virtualizing network functions. The advantages of this approach encompass both hardware and software deployments. However, in this paper, our focus will solely be on the software component, specifically on its cloud cost measurement and optimization. The authors have decided to reserve an in-depth examination of the hardware component for a separate study. Virtualizing network functions provides network operators with enhanced agility and flexibility. Operators can now rapidly deploy and scale software by creating or removing virtual instances as per demand. On the other hand, hardware deployment is also faster due to the use of the COTS model. Network operators use simple and multipurpose hardware. This approach of scaling software and hardware individually significantly improves the responsiveness of network operators to quickly adapt to changing customer needs and market demands. With the virtualization of network functions, network operators can fine-tune their networks to specific business demands and capabilities allowing for diversification among operators. Examples of diversification include scalability and elasticity, vendor choices for hardware, and flexibility to manage software release rate without the rigidity of hardware refresh cycles. This diversification also enables the pace of innovation and empowers operators to respond to market trends and customer preferences in a more nuanced approach, ensuring they can stay competitive in the dynamic telecommunications landscape. Imagine if all network operators were stuck with only vendor-dictated options for both hardware and software, this would impact their level of fine-tuning and the level of diversification in the market they can achieve. Furthermore, the virtualization of network functions offers substantial cost efficiencies since bulk cashflow outlays can be avoided. By decoupling the network functions from dedicated hardware, operators can speed up or slow down software and hardware upgrades separately and at their own comfortable pace. Also, avoiding vendor lock-in, giving network operators a broader range of selection for both software and hardware vendors while avoiding costly upgrade cycles. Network operators also gain a lot of speed and agility specifically in software development and deployment. Virtual instances can be provisioned dynamically, which automatically reduces the overall time for software releases introducing new features or fixing bugs. This does not mean we are excluding other contributing factors to the increase in speeds of software deployments such as the adoption of agile methodologies, the introduction of various new technologies, and hyperscaler’s introducing higher order services. Now that we have seen the benefits of virtualizing network functions, let’s focus on some of the challenges that have been brought in through this model for network operators, specifically in the software component when using virtual instances in either a private or public cloud. The introduction of virtualization brought about a shift in the pace at which software could be deployed when using cloud. However, this advancement also resulted in the removal of certain advantageous constraints that were previously imposed on software deployment due to financial considerations. To gain a more comprehensive understanding of this matter, let us delve further into the traditional approach that was prevalent before the use of the cloud. In the past, software development teams, typically centralized within an IT department, engaged in extensive negotiations with the finance department to obtain the necessary funding for developing new software or upgrading existing releases. These resources, once approved, went through a typically long cycle of identification, purchase, and deployment prior to any software development activities. Software development teams were required to present a compelling case to the finance department, justifying the introduction or modification of a business capability and the requirement of resources for the same. This intricate process, although time-consuming, ensured that the deployment of resources was subject to meticulous evaluation, considering questions like “how does this addition or change in business capability impact the overall strategic value?” However, with the advent of provisioning virtual instances in a cloud for development and deployment purposes, software development teams, and more specifically engineers, gained a considerable degree of autonomy away from finance and justification. They were now empowered to freely provision instances as needed to meet their software requirements, without giving the same degree of consideration to the value associated with each deployment with respect to the business capability or eventually the business activity it is supporting. This predicament was further exacerbated for finance. Suddenly, the traditional capital expenditure associated with resource provisioning began to transform into operational expenditure, requiring the finance department to adapt and prepare for this notable shift. Additionally, the visibility and control that finance teams once had, enabling them to identify the purpose of a provisioned instance for a specific project, became exceedingly challenging to attain from private or public cloud environments. Consequently, when the strategic viability of a project diminishes, finance teams encounter difficulties in enforcing rigorous clean-up or repurposing measures for the associated instances. This lack of visibility and control is prompting finance teams to seek answers regarding where to initiate their efforts, sparking the rise of a new discipline known as FinOps, or financial operations, which has gained substantial traction within the industry. Aptly so, FinOps Foundation executive director J. R. Storment says, noting that the “cloud removes finance from the buying process and hands the credit card to cloud engineers.” This paper partly concentrates on the impact of transitioning from a capital expenditure model to operational expenditure model on finance during software operations, neglecting any discussion regarding the impact on finance during software development. However, it is worth noting that the problem of transitioning from capital expenditures (CapEx) to operational expenditures (OpEx) due to the adoption of cloud also persists during software development, necessitating a comprehensive study on its own. When considering the convergence of two significant trends, the challenges faced by network operators become evident. In a landscape where virtualizing network functions and embracing cloud deployments for enhanced speed, adaptability, and innovation are standard practices, the questions arise: 1. How can a network operator effectively manage and control the software operational expenses that accumulate? 2. What strategic value can network operators add to their portfolio by being on top of such operational expenses? 3. Is all this effort to go after these expenses even worth the time & effort? In this paper, the authors dedicate some amount of time to discuss all the issues there are in cloud cost measurement, interpretation, and optimization in a cloud environment and suggest some strategies on how best to get a hold on the situation to avoid overrun of operational expense.