CIFC adopts agile Azure platform for mission-critical credit services

A people-first approach to cloud migration

An alternative credit specialist, New York–based CIFC employs more than 150 professionals in the United States while serving institutional investors globally. CIFC saw an opportunity to modernize its aging IT infrastructure with a move to cloud services. Many enterprises make this move knowing that they can exchange the operational expenditures (OpEx) associated with underutilized datacenters for the efficiency of cloud-style, pay-per-use capital expenditures (CapEx). But to stay competitive in an industry rife with digital disruptors, CIFC went a step further—not just modernizing its technology stack but also embracing cloud-native architectures.

According to CIFC Chief Technology Officer Adrian Iosifescu, “Moving the CIFC Asset Management technology ecosystem to the cloud has been an essential component of our technology strategy.” A cloud platform offers the elasticity, scalability, and flexibility needed to support the company’s rapid business development and diversification.

The firm asked DataArt for assistance. As a Microsoft Gold Certified Partner, DataArt brought 20 years of enterprise IT consulting expertise to the CIFC project. Phase one of CIFC’s cloud journey began with the rapid retirement of the existing datacenter in an Azure lift-and-shift, but it didn’t stop there. To take full advantage of the performance, scale, and cost benefits that cloud hosting provides, shifted workloads must be optimized. For CIFC, that meant working with DataArt to refactor monolithic application architectures using modern microservices and containers.

CIFC’s Iosifescu acknowledges the power of the partnership, saying “DataArt brought to bear expertise, knowledge, and qualified resources to help us realize the goals of this and other projects. They also have been our venue into Microsoft.”

From lift-and-shift to cloud native

The CIFC and DataArt teams started with the Analytics Core application. This number-crunching engine builds targeted valuation models that meet strategic business needs. Analytics Core processes high volumes of data in scheduled batches and on demand, consuming massive computing resources in the process. Portfolio managers use the results to make well-informed investment decisions. But before the migration, getting results meant waiting overnight for batch runs to complete.

CIFC developed a prototype that ran on a virtual machine (VM) as a proof of concept, but it wasn’t designed for production scale. The next step was to use Java microservices—a more agile architectural approach that is better able to meet spikes in traffic. To simplify the deployment and management of a microservices-based architecture, the team chose AKS as the hosting platform. AKS provides fully managed Kubernetes clusters that help developers streamline the deployment, scaling, and management of containerized applications.

As Iosifescu relates, “To really take advantage of the cloud technology, we initiated phase two of the project: have the application leverage Azure-native capabilities. Here is where the real benefits shined. We saw improved performance, true elasticity of using resources, and believe it or not, a significant cost reduction.”

The cost benefits are no accident. AKS offers elastic provisioning of capacity without the need to manage the infrastructure. CIFC pays for only the VMs, along with the associated storage and networking resources that are consumed. Given the Analytics Core application’s intensive processing requirements, it was the perfect candidate for a cloud-native makeover, observes Yury Zaryadov, DataArt Vice President, Enterprise Solutions. “The move also provides a future-proof solution, considering the rapid growth of financial data volumes that need to be processed in the company.”

A supercharged microservices-based architecture

To rebuild Analytics Core as a Java calculation engine, the engineers used a combination of open-source technologies and Azure services. The Analytics Core application includes a user portal, a command-line interface, and web APIs, but at its center is a calculation engine used for accurate cash-flow projections and price/yield analytics.

Using the Spring Framework, they created an architecture based on Spring Boot. Microservices run in containers, giving engineers the flexibility to build and run containers on any development machine, using their tools of choice.

Engineers store and share the container images through Azure Container Registry, a fully managed, geo-replicated registry for Docker and Open Container Initiative images. The calculation engine runs in five main modules, together with two message queues that coordinate the tasks and results via RabbitMQ, an open-source message-broker software.

According to Zaryadov, “With the implementation of the AKS cluster, we achieved great progress. The migration successfully addressed performance and scalability issues of the system.”

Dependencies and databases

In moving the Analytics Core data tier to Azure, the engineers faced several challenges. Strong dependencies tied the application logic to an existing data warehouse solution and an internal analytics database, and an inefficient data layer design caused database deadlocks and blocks. Moreover, the database support costs were growing.

The new application architecture minimizes component dependencies by isolating application logic from the databases. It was a good start, but the engineers wanted more.

An early version of the architecture deployed Microsoft SQL Server on VMs. To simplify system maintenance and optimize the application, the team switched to Azure SQL Managed Instance, a fully managed, always up-to-date SQL instance in the cloud. It stores the cash-flow models used by the calculation engine.

Calculation results are stored in MongoDB. To determine which database worked best for the calculations, the engineers compared query performance in each. SQL Server outperformed MongoDB by a wide margin, making it the clear choice. This data is consumed by a REST API that serves the front-end dashboard. Users have quick access to the calculations they need.

According to Alla Lemlekh, DataArt Senior Vice President of Strategic Partnerships, “It was a challenge to store, process, and analyze such a significant amount of financial data,” but the migration was a success. “Overall enhancements and modernizations done for the system allowed it to perform better and deliver faster analysis, providing even more impact on CIFC’s business and IT solutions,” she adds.

Optimization applies to business processes and to data. CIFC development teams need to work independently but stay connected. The entire deployment is automated using a continuous integration and continuous deployment (CI/CD) process. Azure DevOps Services and Git repos hosted in Azure Repos serve as CIFC’s source control and deployment tools, replacing Subversion and Octopus Deploy. Azure Artifacts, a feature of Azure DevOps Services, enables CIFC to incorporate Apache Maven, the team’s preferred build automation tool.

The DevOps improvements extend to database deployment, something that, in the past, CIFC handled manually. Using SQL Server Data Tools (SSDT), the engineers can now manage the database schema as a project, work in an SSDT repo, and use GitHub flow to make changes that colleagues can review and approve.

The effort pays off in predictable database deployment, stable releases, and greater visibility into application performance and container health.

Lower costs, faster performance, and greater elasticity

Speeding up the Analytics Core application has been a huge benefit for the company, as Iosifescu relates. “The most impactful result with this particular application was the ability to reduce the execution of these complex analysis scenarios from overnight to almost real-time,” he notes. “This presented our business with a tremendous opportunity to make intelligent business decisions on the spot.”

The move to Azure gave CIFC the modern cloud platform it needs to face the next challenge. As Lemlekh sees it, “The technologies and services provided by Azure expanded CIFC projects’ capabilities—not only improving performance and scalability but also reducing maintenance overhead and even unlocking new functionalities for the business.”

With more modernization projects in the queue, CIFC continues to work with DataArt and take advantage of Azure Managed Services. Up next: the migration of a data warehouse and the Securities Master, the firm’s organization-wide database that stores fundamental, pricing, and transactional data for its financial instruments. The firm is also exploring how Azure Machine Learning and natural language processing technologies can improve analyses.

“We had three criteria to measure success,” Iosifescu concludes. “Improved performance, true elasticity, and overall cost. The move to Azure was a success from all three points of view.”