A Performance Evaluation Method for Climate Coupled Models

In the High Performance Computing context, the performance evaluation of a parallel algorithm is carried out mainly by considering the elapsed time for running the parallel application with both different number of cores and different problem sizes (for scaled speedup). Typically, parallel applications embed mechanisms to efficiently use the allocated resources, guaranteeing for example a good load balancing and reducing the parallel overhead. Unfortunately, this assumption is not true for coupled models. These models were born from the coupling of stand-alone climate models. The component models are developed independently from each other and they follow different development roadmaps. Moreover, they are characterized by different levels of parallelization as well as different requirements in terms of workload and they have their own scalability curve. Considering a coupled model as a single parallel application, we can note the lacking of a policy useful to balance the computational load on the available resources. This work tries to address the issues related to the performance evaluation of a coupled model as well as answering the following questions: once a given number of processors has been allocated for the whole coupled model, how does the run have to be configured in order to balance the workload? How many processors must be assigned to each of the component models? The methodology here described has been applied to evaluate the scalability of the CMCC-MED coupled model designed by the ANS Division of the CMCC. The evaluation has been carried out on two different computational architectures: a scalar cluster, based on IBM Power6 processors, and a vector cluster, based on NEC-SX9 processors.