Numerix Scales HPC Workloads for Price and Risk Modeling Using AWS Batch

Numerix provides its analytics software to more than 250 global clients, including banks, regulators, and insurance companies. Its extensive mathematical models price deals against a wide variety of market states to simulate the likely effects if stock prices took a tumble. Financial institutions rely on this data to make decisions with billion-dollar implications, and they require the most advanced analytics available. Further complicating matters, financial markets have been in unprecedented territory since the early days of the COVID-19 pandemic. Trade and risk management information is especially valuable in this time of instability. “We have clients that are doing portfolios of 20,000 trades,” says Jockle. And those portfolios are only growing larger as firms embrace risk analytics in an attempt to shield themselves from vulnerability.

The complexity of this increase in trading and analytics volume is an immense mathematical challenge that requires a lot of compute power. Bill Humphrey, chief technology officer at Numerix, says, “For clients to run our solutions on premises, we have to tell them, ‘This is how many CPU cores you need to have in your data center when you install our software and run it every day. And you’ll have to buy even more next year because your portfolio is growing.’” That startup cost has been a barrier to the adoption of Numerix tools.

“The cloud has been an inevitable journey for Numerix to provide efficiency and availability,” says Jockle. Numerix began undertaking some software-as-a-service projects in the cloud in 2012. In 2019, the migration to AWS accelerated as engineers started using Amazon Elastic Compute Cloud (Amazon EC2), which provides secure and resizable compute capacity in the cloud, to run its HPC solutions. Numerix started using AWS Batch in 2021 to dynamically provision the optimal quantity and type of compute resources on Amazon EC2. With the new approach, analytics performance has improved by 180 times.

Numerix leaders agree that adopting cloud-native orchestrator and serverless architecture has been the key to taking advantage of the full elasticity of the cloud. Although Numerix used a lift-and-shift approach in the early stages of the migration, the full migration to a serverless model was a milestone. “The serverless model is exactly what we need so that we don’t have expensive resources running all the time,” says Humphrey. “We submit these workloads to AWS Batch, which orchestrates compute resources by provisioning the right Amazon EC2 instances for the jobs submitted, runs these jobs, and then shuts the instances down when the work is completed, and we’re charged for only the actual seconds of use.” Numerix uses AWS Step Functions, a low-code, visual workflow service for modern applications, to run its serverless capabilities.

More importantly, by using dynamic resource allocation on AWS, Numerix can meet demanding client constraints more effectively. “Using AWS Batch, we meet service-level agreements of 40 minutes or less on portfolios with tens of thousands of trades,” says Jockle. “That’s absolutely unheard of.” Engineers are staging information using Amazon Simple Storage Service (Amazon S3), cloud object storage built to retrieve any amount of data from anywhere. The increased memory and storage capacity on AWS have reduced bottlenecks across the analytics process. Now, Numerix is much better prepared to take on larger portfolios. Instead of telling clients that they will have to wait several months to purchase, receive, and install servers each time they scale up, Numerix can help them respond to sizing changes in days or hours. “Just being able to adapt quickly is a huge win,” says Humphrey.

Many of Numerix’s clients have appreciated the transition to a cloud-first mindset. “In the cloud model, clients no longer need a very large IT department to run our HPC solutions,” Humphrey says. Instead of buying more servers every time they scale up, organizations can adapt to sizing changes in the cloud in a matter of hours. Numerix also makes extensive use of Amazon EC2 Spot Instances, which help users to run fault-tolerant workloads for up to 90 percent off Amazon EC2 On-Demand pricing. By using Amazon EC2 Spot Instances and serverless technology, Numerix has experienced significant cost savings.

Numerix is eager to transition more of its clients to the cloud and is working to expand its software-as-a-service model as a key delivery and operational framework. “AWS provides such a huge range of services and capabilities,” says Humphrey. Instead of preparing hardware for the worst possible case, clients pay for computing power as they go.

These technical enhancements have a real-world impact. “Our clients are using our risk analytics to avoid billion-dollar losses,” says Jockle. “The introduction of near-real-time analytics with the virtually limitless scalability of AWS has been a real game changer.”