CONTEXT
SG CIB SGXIB is currently using a 5000 cores Compute Grid for its pre-trade (read: real-time) CVA calculation. Those resources are turned-on permanently – which costs a lot to the firm. The objective of this Cloud Computing project was to reduce the environmental footprint and the operational costs.
SOLUTION
Leveraging AZURE Cloud from Microsoft, the application was rewritten in C# Using the native Azure scheduler Turn on/off nodes when needed (up to 10000 cores)
RESULT
Conception and development of a new HPC-Cloud-native application. Used daily in production over 10,000 cores. Benchmark was done up to 30k cores. A distributed orchestration workflow manager was built to support the applications More than 200 users worldwide