Automatic tuning for a systemic model of banking originated losses (SYMBOL) tool on multicore

Economy is a complex machine and the same holds for financial markets. Though mathematical and statistical models aim to represent only a simplified version of such complexity, in the latest years the became more and more sophisticated. This fits in a more general framework where scientific applications are developed with more and more complexity and accuracy. However, these precisions and complexities have brought, as result, that applications need more computational power in order to be executed in a reasonable amount of time. In this sense, the multicore environments are playing an important role to improve and to optimize the execution time. These environments allow us to include more parallelism inside the node. However, to take advantage of this parallelism is not an easy task, because we have to deal with some problems such as: cores communications, data locality, memory sizes (cache and RAM), synchronizations, data dependencies on the model, etc. These issues are becoming more important when we wish to improve applications’ performance and scalability. This paper describes an optimization method developed for SYMBOL model, which is based on analyzing the application’s weakness in order to exploit the advantages of the multi-core architecture. SYMBOL is a serial micro-funded statistical tool , which analyzes the consequences of bank failures, taking into account the safety net, such as deposit guarantee schemes or resolution funds. However, this tool has two main weakness that make execution times grows considerably: the first one is related to the execution with large input data (e.g. large banking systems) and second is when we are trying to scale the number of default scenarios. Hence, our method defines a set of steps, which allow us not only to include the optimization techniques, but it also permits us to select the best amount of cores in order to take advantages of the multicore. All these improvements are done in an automatic and transparent manner with the aim of improving the performance metrics of SYMBOL tool. Finally, experimental evaluations show the effectiveness of our new optimized version, in which we have achieved a considerable improvement on the execution time. Comparing the original serial version of SYMBOL with the automatic parallel one the execution time has been reduced around 96% in the best case tested. Moreover, SYMBOL simulations are automatically tuned using the input parameters allowing to make better use of the computational machine.

MURESANO CACERES Ronal Roberto;  PAGANO Andrea; 

2014-11-25

World Academy of Science, Engineering and Technology

JRC91352

1307-6892,