How cloud computing transforms actuarial modeling infrastructure

Large actuarial modeling operations were traditionally hosted on-premises. The arrival of cloud-managed created a compelling reason to evaluate the cost-benefit of where to host modeling technology infrastructure. The industry has seen a shift of migrating modeling infrastructure to the cloud from what used to largely be on-premises solutions to a variety of new platforms. Adding to the choices, many actuarial software vendors now offer cloud-managed solutions.

To better understand the opportunity, we’ll begin with a high-level overview of the technology infrastructure that supports modeling operations under both on-premises and cloud-managed solutions. This overview will provide the reader with a better understanding of the infrastructure required and how procuring and managing it compares for on-premises and cloud-managed solutions.

Figure 2: Summary of modeling technology infrastructure core components

As shown in Figure 2, there are several core components to modeling infrastructure. A detailed discussion is beyond the scope of this article, but the summary below describes the core components that underly both on-premises and cloud-managed solutions.

Core components of technology infrastructure required for modeling operations include:

1. Networking – enables computing and communication among users, services, applications and processes through hardware and software, including routers, switches, network operations software, security and IP addresses.
2. Servers – dedicated machines that serve information to other clients across a network, and may be designed for specific tasks such as file servers, application servers, database servers, compute/job servers (“workers”), etc.
3. Data storage – the storage or databases that hold information for the applications, providing services like updating, deleting and finding data, and performing searches across data.
4. Compute processing – typically thought of as one or more job servers (“workers”) with a set of CPU/GPU processors that perform jobs in a queue. These servers typically come with a sizeable number of CPU cores that perform the computations.
5. Operating System (OS) – is software required to run applications and utilities, acting as a bridge between application programs and hardware of the computer and network.
6. Applications – are software packages that perform specific functions. Common applications here will be the vendor-based modeling applications (e.g., FIS Prophet, Moody’s Axis, etc.) and supporting applications like Excel, PowerBI, etc., that actuaries typically interact with for analysis.

Options for procuring modeling infrastructure

For on-premises solutions, an organization would procure its own infrastructure (as described above), locating it on the premises of the organization’s data center (or similar) rather than acquiring through a service provider. The modeling infrastructure and the applications are under the ownership of the company and not “rented.”

In contrast, companies can explore a variety of cloud-managed solutions. Cloud-managed is a method of enabling on-demand network access to a shared pool of resources, including networks, servers, storage, applications and services as described above.

Modern cloud computing is often categorized into three categories:

• Infrastructure as a Service (IaaS): refers to the hardware and software that act as the foundation to support applications and operating systems, including servers, storage, networks and virtualization
• Platform as a Service (PaaS): tools and services designed to make coding and deploying those applications quicker and efficient without worrying about infrastructure provision or the operating system
• Software as a Service (SaaS): applications designed for end-users and delivered over the web, providing needed infrastructure, platforms and applications as a service that requires only configuration.

To summarize, IaaS offers networking, storage, servers and virtualization on demand. It is the most fundamental level and provides a cloud-based foundation to build on top the operating systems, data and applications. PaaS then layers on an organization’s operating systems. This offers an environment that is ready for development and deployment. Finally, SaaS incorporates an organization’s data and applications, or software, which sit on the top of the technology stack.

Figure 3: Understanding infrastructure options

As shown in Figure 3, the range of options from on-premises to cloud-managed varies by what you manage vs. the service provider manages. IaaS/PaaS and SaaS are cloud-managed options. By better understanding the options for procuring infrastructure, we can now better understand and discuss the comparisons of on-premises to cloud-managed.

As mentioned above, the industry has seen a shift from on-premises to cloud-managed. However, cloud-managed operations may not always be cheaper than on-premises, and flexibility of hosting arrangements can vary, hence it’s crucial to understand options, costs and trade-offs. The following table summarizes the comparisons of on-premises and cloud-managed.

Table 1: Comparison between on-premises and cloud-managed solutions across technology, process, people and costs

To help understand current adoption trends and costs associated with cloud computing supporting actuarial modeling use cases, Ernst & Young LLP conducted a survey in 2023 of 20 different insurance companies, which included a diverse mix of type (stock and mutual), size and actuarial software usage. The goal was to understand the industry practice of hosting (e.g., on-premises, cloud-managed), utilization of distributed processing capacity, computing cost and expectations about future cloud usage.

Some of the key findings from the survey include: