MG-ALFA takes to the cloud: High-performance computing for Solvency II
With the advent of Solvency II in Europe and impending Solvency II-style regulation in the United States, Asia, and other regions, insurance companies find themselves in need of more powerful analytical tools than ever before. The modeling techniques dictated by Solvency II are extremely computer-intensive and timelines for reporting are much tighter, which means that users of today’s desktop or client-server systems will find it extremely challenging to obtain results with the required precision and timeliness. Additionally, enhanced operational and governance requirements can ensure more stringent operational and governance processes to allow companies to explain and demonstrate why they made the choices they did and document each step in an auditable way.
All of these factors together create significant challenges for insurers who need to implement the analytical tools and processes to meet these requirements. In collaboration with The Phoenix Group, the largest consolidator of closed life funds in the U.K., Milliman has created a revolutionary cloud-based version of its industry-leading MG-ALFA® system that meets the requirements of insurers facing the need for modeling and analytics tools orders of magnitude more powerful than ever before. Because it is based on the Microsoft Azure platform, users have access to virtually unlimited high-performance computing resources on demand at a small fraction of the cost of owning and operating those resources. This article details Milliman’s award-winning work in harnessing the power of the Microsoft Azure grid computing solution to an analytical platform—the MG-ALFA cloud system—specifically designed to meet the needs of insurers under contemporary regulatory schemes such as Solvency II.
Actuarial modeling in the cloud
MG-ALFA is one of the most widely used actuarial modeling tools in the world. It is used for financial reporting, risk management, asset-liability management, planning and forecasting, and product development. It is a complex and extremely powerful tool, with the ability to distribute the calculations across hundreds of processors using Milliman’s highly scalable grid computing capabilities. So what value does moving to the cloud provide? Because MG-ALFA can now provide on-demand, unlimited computing resources in a controlled and highly automated environment, users can have increased confidence in the accuracy and timeliness of results, not to mention more control over the associated costs.
To date, many insurers have invested in large, in-house grids to support the actuarial modeling requirements. The challenge with this approach is that the complexity of the calculations required by Solvency II and the tight reporting timelines would require an extremely large grid for short periods. The expense of building and maintaining this infrastructure, especially when it is under-utilized for much of the time, makes this option difficult to justify for many insurers.
In cloud computing, also referred to as “software as a service” or SaaS, servers and software are hosted and managed by a service provider and accessed over a secure Internet connection. While cloud computing has become something of a buzzword lately, we believe it will be of great value to insurers under principles-based regulations.
Specifically, cloud computing can provide utility-like access to truly large-scale networked computing resources, meaning you pay only for what you use. It therefore eliminates the large capital expenditures necessary to deploy such resources on-premises, as well as the cost and complexity of managing, upgrading, and supporting the infrastructure. Because it is accessed over a secure Internet connection via a browser, it can be accessed from virtually anywhere, on any Internet-enabled device from laptops to smartphones to tablets. This enables actuaries to perform work from more places than ever before.
With the MG-ALFA cloud system, we started from the existing power of the MG-ALFA desktop application. We then developed extensive proprietary computing infrastructure on top of the Microsoft Azure cloud computing service. To date, we have scaled to more than 8,000 compute nodes at a time—a supercomputer level of processing power available practically at the touch of a button.
Certainly, security is a significant concern for financial services companies. That is one of the reasons for the choice of Microsoft Azure, as it is an enterprise-grade platform with attendant security and reliability. Additionally, because Microsoft has Azure data centers worldwide, data can be hosted in jurisdictions appropriate to the client.
One of the first considerations in designing a cloud-based analytical tool is how to get data into the system. The MG-ALFA cloud system is designed to interface with external integration points that allow it to accept input, report its state to other systems, request input, send results, and report when jobs begin, run, and end. System integrators will be able to write external software that provides input to the MG-ALFA cloud system for model points, economic scenarios, and assumptions. This enables full automation of the process, eliminating human error and time associated with the manual updating of models.
Versioning and auditing
Prior to Solvency II and similar regulations, running complex stochastic financial projections was a relatively ad-hoc process, primarily supporting internal reporting. Using these complex models for regulatory reporting creates a need for a very controlled environment and processes that provide documented, reproducible, and auditable results.
Tracking this information manually in a complex enterprise environment is virtually impossible. To meet the need for enhanced auditability and make the process of developing, running, and understanding models more efficient, the MG-ALFA cloud system has been built to allow for a robust version tracking system.
The ability to make changes in a production environment is tightly controlled. All jobs have tracking information to indicate exactly which data was used in the job submission so results can be traced back to the input that produced them.
Practically speaking, it will be possible to track which changes were made to any data object, while also viewing when the change was made and by whom. Additionally, it will be possible to run multiple versions of a data object and compare the results to understand the impact of changes.
It is important to note that in an on-premises system, enterprise-grade storage would need to be deployed in-house to store all of this data. With the MG-ALFA cloud system, you pay for the storage you need without the costs of deploying, managing, and upgrading hardware.
Another challenge in existing actuarial modeling environments is creating and maintaining models across the entire organization that use consistent logic and assumptions. In the current desktop application environment, it is necessary to have copies of models to distribute the workflow. This leads to duplication of work, inconsistency, and modeling error. Moving to the cloud will allow for multi-user capabilities, enabling a highly collaborative environment.
Workflow management and automation
Solvency II and similar regulations combine vastly increased model complexity and hundreds of required runs with strict time limits to obtain results, creating a high-pressure environment to say the least. With such a short window to obtain results, you simply can’t risk someone getting sick or taking a long lunch and losing a day of processing time, not to mention someone making unaudited changes to the model. This situation requires a repeatable process than can happen every quarter without any human intervention.
The MG-ALFA cloud system provides workflow management and automation capabilities that greatly streamline the process and reduce hands-on time to a minimum.
- System integrators will be able to attach to extension points in the MG-ALFA cloud system that allow them to control the workflow of users, and start workflows on behalf of external systems.
- Workflows will allow the system to perform actions such as notifying users when assumptions are due to be updated, providing gated changes and job submissions that require peer review or management approval, and so on.
- Monthly, quarterly, and ad-hoc valuations can be run in an almost entirely automated manner.
Service management and business continuity
As a business-critical SaaS platform, the MG-ALFA cloud system employs numerous measures to maintain high availability. Milliman staff provides 24x7 monitoring of services using proprietary diagnostics, event management, and other service management techniques. Using built-in replication functionality in Microsoft Azure, MG-ALFA data is replicated across Microsoft data centers with the ability of the MG-ALFA cloud system to recover from individual data center service interruptions.
Milliman was recognized as the Microsoft Technical and High Performance Computing Partner of the Year for our work on the MG-ALFA cloud system with The Phoenix Group, and we plan to make the system commercially available soon. We are excited to take advantage of the latest cloud computing technology to deliver a cost-effective solution for our clients who are facing ever more complex valuation challenges.
Milliman STAR Solutions–Vega