Hazelcast, the fast cloud application platform, is releasing a reference implementation that simplifies a financial service organization’s ability to execute and scale financial risk calculations in the cloud while gaining real-time performance and fully-utilizing the resource-heavy investment.
Additionally, Hazelcast is also announcing the availability of its managed service, Hazelcast Cloud Enterprise, on Google Cloud Platform (GCP), which was used as the environment for this reference implementation.
Credit value adjustment (CVA) is the financial measure of risk for a given set of trades should the counterparty default on its obligations. The Hazelcast reference architecture showcases how an in-memory computing platform can leverage the advantages of cloud environments, especially around elasticity and agility.
CVA calculations are large-scale computational jobs, and the ability to scale on-demand lets financial services firms control their computing costs by only paying for the resources they use.
“The economic benefits of the cloud are well-noted, especially around the ability to only pay for the resources you consume. On-premises deployments that are dedicated to specific workloads do not fully utilize the available resources, so you do not get full value on your infrastructure investment,” said John DesJardins, chief technology officer (CTO) at Hazelcast.
“For use cases where latency, predictability and scale are critical, such as this risk management scenario, Hazelcast is providing the financial services industry with a blueprint for migrating mission-critical applications to the cloud and using ultra-fast in-memory technologies to gain a competitive advantage.”
Once in-memory technologies are implemented, trades can be loaded from a database and distributed across a caching layer of multiple nodes. The input can then be fed into a compute grid to run blocks of calculations in parallel, capturing the results back to more distributed caches. In this scenario, a large server farm of around 40,000 cores is often required to complete the massive number of calculations in minimal time.
Hazelcast distributed data store with data-aware stream processing takes this model from a batch method to a pipeline approach that pushes input data to the calculation farm, orchestrates the computations efficiently and then aggregates the results as they are produced.
The “straight-through processing” approach eliminates data writes at each stage of the computation, thus further accelerating the overall speed of the calculation. The benefit is that customers can move from daily or maybe hourly, to enabling ad hoc calculations based on any events that might change market dynamics.
This evolution allows financial institutions to react more quickly to changing market conditions, whether those are triggered by business news, a national disaster, geopolitical events or other factors. Cost-effective risk management benefits not only financial institutions but the broader markets and global economy.
When deploying this implementation in the cloud, as it was in GCP for this demonstration, companies are easily able to scale infrastructure to adjust to the volume of calculations or required runtime.
Demonstrating this cloud-native approach, Hazelcast scaled its usage of GCP to a cluster of 300 containerized Hazelcast nodes, running across a Kubernetes cluster of 110 virtual machine instances based on Intel Cascade Lake hardware. With a traditional on-premises approach, a large number of machines would have to be purchased, forcing the customer to make an economic compromise that selects a machine count somewhere between peak load and low load counts, being ideal for neither.
Hazelcast Cloud Enterprise
Hazelcast Cloud Enterprise is a managed service where Hazelcast experts maintain the software deployment on behalf of customers, freeing employees to focus on developing business-critical applications to improve real-time insights. With its availability on GCP, Hazelcast Cloud Enterprise is now available on all major clouds, including Amazon Web Services (AWS), Microsoft Azure and IBM Cloud.
Featuring built-in WAN Replication, Hazelcast Cloud Enterprise can efficiently replicate data from any location to another, ensuring the availability of an application and its data. This multi-cloud strategy enables a broader range of geographic data distribution to run applications and data closer to the end-users, enabling a greater level of availability by providing active replica clusters across zones, regions, cloud providers, and on-premises sites. Developers can thus build business applications on the same application framework no matter where the application is deployed, even at the edge or in hybrid cloud environments.