Organization of calculations in a hybrid high-performance computing cluster for parallel execution of heterogeneous tasks

Approaches to provision of computing resources for solving applied and fundamental scientific problems are considered. The article suggests methods and algorithms for adapting and executing user applications on a hybrid computing cluster that uses specialized accelerators to perform calculations. The authors offer solutions for migrating code in high-level languages to the target platform of a hybrid computing cluster, algorithms for providing computing resources in a multitask environment, and methods for creating an adapted execution environment using virtualization technology.