GLOBE Claritas Cluster Support
GLOBE Claritas provides a highly scalable solution for running on Linux clusters.
With our approach, you can split almost any CPU-intensive processing flow across multiple instances and cores; you simply flag the section of the flow that needs to run in parallel and indicate how many cores you need to use.
These parallel flows can include modules that you have developed in house – whether they are fully implemented inside the Claritas API, or just left as calls from the RUNPYTHON or TRFUDGE modules.
We’ve also developed “stand alone” parallel applications for some of the more CPU intensive tasks, such as pre-stack time migration and SRME modelling. This suite of applications uses a common suite of routines that are engineered to allow us to optimise them for parallel use.
Initial benchmarking of the parallel capability utilised the New Zealand Super Computer service, but recent tests on the Amazon EC2 cloud have proved that our process scale out to many hundreds of cores and beyond.