Parametric Optimization on HPC Clusters with Geneva

Abstract

Many challenges of today's science are parametric optimization problems that are extremely complex and computationally intensive to calculate. At the same time, the hardware for high-performance computing is becoming increasingly powerful. Geneva is a framework for parallel optimization of large-scale problems with highly nonlinear quality surfaces in grid and cloud environments. To harness the immense computing power of high-performance computing clusters, we have developed a new networking component for Geneva, the so-called MPI Consumer, which makes Geneva suitable for HPC. Geneva is most prominent for its evolutionary algorithm, which requires repeatedly evaluating a user-defined cost function. The MPI Consumer parallelizes the computation of the candidate solutions' cost functions by sending them to remote cluster nodes. By using an advanced multithreading mechanism on the master node and by using asynchronous requests on the worker nodes, the MPI Consumer is highly scalable. Additionally, it provides fault tolerance, which is usually not the case for MPI programs but becomes increasingly important for HPC. Moreover, the MPI Consumer provides a framework for the intuitive implementation of fine-grained parallelization of the cost function. Since the MPI Consumer conforms to the standard paradigm of HPC programs, it vastly improves Geneva's user-friendliness on HPC clusters. This article gives insight into Geneva's general system architecture and the system design of the MPI Consumer as well as the underlying concepts. Geneva, including the novel MPI Consumer, is publicly available as an open source project on GitHub and is currently used for fundamental physics research at GSI in Darmstadt, Germany.

Speeding up Science Through Parametric Optimization on HPC Clusters

“…Looking back at Figure 3, we can see that the performance gain of up to 20% accounts for a large part of the improvements that the MPI consumer achieved compared to the Boost.Beast consumer. Moreover, we consider it positive that the feature has a stronger impact for shorter cost function evaluation times, since the system already runs with near-to-perfect when using longer cost functions, as shown in our previous paper in Section 5.2 [3].…”

“…The latest network component, the MPI Consumer, which we have developed in 2022, improves the usability of Geneva for high-performance computing and allows for fine-grained user-defined parallelization of the cost function. We have already shown the results of the performance tests for this consumer in our last paper [3]. Figure 3 illustrates the relative performance improvement provided by the MPI Consumer over the Boost.Beast Consumer.…”

“…Recently, we have developed the MPI Consumer, a new network component for the Geneva optimization library, which increases its user experience for high-performance computing and also adds the new feature of fine-grained parallelization. We have discussed the design decisions of the MPI Consumer and also shown first performance test results on a 128-core machine in the previous publication [3]. Since this paper builds on top of our last paper, it is recommended to read the last paper before reading this one.…”

“…In our previous work [3], we have shown a plot of a series of tests measuring the performance of the MPI Consumer with different numbers of clients and different durations of the user-defined cost function. Now, we have repeated these tests with the same configuration [3, p. 13] for the two alternative network components of the Geneva library, which have been developed earlier.…”

“…In our previous publication [3], we explained the system design of the MPI Consumer in detail. The two features (1) asynchronous client requests and (2) multithreading have been implemented specifically to achieve great scalability.…”