User story: Société Générale & Qarnot

Société Générale succeeded in reducing its financial simulations’ cost by 50% and its carbon footprint.

With this solution, developers can focus on business needs and rely on the platform for resiliency, load balancing or scalability.
― S. Bonniez, Project Manager at Société Générale

Société Générale provides financial services to 31 million individuals and professionals worldwide, placing innovation and digital technology at the heart of its activities. Its corporate and investment banking business – SG CIB – offers global access to markets via solutions in equities, fixed income & currencies, commodities and alternative investments. SG CIB is committed to supporting corporates, financial institutions and investors by providing advisory, investment, and risk management solutions.

 

Being a responsible company lies at the very heart of Société Générale business lines’ mission and its commitment to a greener and more inclusive economy has been accelerating since the 2000s. Société Générale has partenered with Qarnot Computing and Microsoft Azure to build its new financial simulation platform. Running market activities requires significant amounts of financial simulations running on large scale grid computing infrastructures. The new platform is flexible, scalable and environmentally responsible.

A scalable, resilient and simple platform

Financial simulations are computationally intensive: they  involve several thousands calculation tasks, taking from a few seconds to several minutes each and rely on a lot of data.

 

Since Société Générale has offices all around the world, that means that jobs can be triggered at any time, any day. Some of the simulations also have strong computation time constraints.

 

We designed a solution that:

  • Exposes a simple REST API to client applications within Société Générale.
  • Handles calculation jobs ranging from a few tasks to several thousands.
  • Provides caching of financial data for efficient dispatching of tasks.
  • Scales with the number of jobs and tasks.
  • Is available 24/7.

 

All this is achieved in a context where delivery must happen often (simulation libraries evolve continuously) and infrastructure costs must be kept as low as possible, although thousands of CPUs may be required to perform some simulations.

 

The key components of that solution are the following:

  • A https web gateway exposing simulation services as a REST API.
  • A collection of micro-services handling data caching and the orchestration of simulation jobs, from the dispatching of tasks to the retrieval of the results.
  • Several grid computing providers. We currently target Azure Batch and Qarnot Computing’s platform, but new providers can be added very easily.
  • The web gateway and the micro-services are native Service Fabric applications, all deployed in a scalable cluster in Azure Cloud.

 

Being able to target multiple cloud providers is a guarantee that a job will always find room to run at the best possible price.

Netork architecture
The challenge was to deliver a modular, scalable and resilient application in a very short timeframe.

The benefits

With Service Fabric, Qarnot Computing and Société Générale have developed an innovative yet high-quality solution with just the right amount of effort. Now that the simulation platform is in production, the team focuses on integrating new types of simulations and scaling the platform to handle them so that more and more client applications can move away from legacy systems.

HPC

A question? A request for proposal?

Monte Carlo Simulations, actuarial sciences, risk management, stress-tests… Qarnot computing, European, ecological, secure and economical cloud provider, has developed a strong and valuable expertise in risk calculations since 2010.

They trust us

The team focuses on integrating new types of simulations and scaling the platform to handle them so that more and more client applications can move away from legacy systems