Advances of the computing industry in SOA infrastructure have made service oriented applications practically attainable in the field of high performance computing for earth sciences. We have applied SOA to modeling and inversion of geophysical well logs. Parallelized log simulators are accessible as a Web service to a variety of interpretation workflows. We have developed an HPC management infrastructure based on SOAP. It is a fit-for-purpose system with a low runtime overhead and a novel flexible job submission mechanism, suitable for applications running on multi-cluster environments. We describe the system architecture and discuss performance benchmarks, including a novel way of estimating performance on heterogeneous clusters.