GPFS Performance Tuning – Oil & Gas Use Case
GPFS is a hugely popular file system for many applications today, including Bioinformatics, Seismic Data, Financial Modeling, and Genomic Research, among many others. Fundamentally, GPFS allows a group of computers concurrent access to a common set of file data over a SAN infrastructure, a TCP/IP network or a combination of connection types. Because the system has so many features and “levers” that can be activated depending on the application, it’s important to map ongoing needs and priorities with an experienced partner to get the best bang for your application’s buck.
GPFS Configurations
GPFS is a base offering but it has a tremendous amount of additional functionality including AFM, ILM, HSM, as well as different variations on the architecture. GPFS provides storage management, centralized administration and shared access to file systems from remote clusters providing a global namespace. These clusters can be made up of anything from a single node with tiered storage to thousands of nodes, depending on the application’s needs. For example, supercomputers used for weather modeling, complex simulations, and predictive modeling often exceed hundreds of thousands of cores.
As application workloads vary over time and over users and software, there is not a single GPFS configuration that is best for everything. GPFS performance tuning and configuration is largely iterative and depends on changing workloads and priorities. The interactive process of data collection and customization informs which features should be utilized and optimized on an ongoing basis.
GPFS Performance Tuning for Oil & Gas
Leveraging GPFS, our team has created custom solutions for many verticals. When one of the largest seismic research companies in the world was in the process of a tech refresh for their data storage, for example, we recommended a building block approach that was fine tuned to their existing environment which included 5,000 compute nodes. We then tuned GPFS performance for their existing networking infrastructure. The client soon rolled out over 12 PB of storage using the GPFS solution we developed.
See why and how GPFS is one of the premier parallel file system architectures in use in HPC clusters around the world. Get in touch with our team to put it to work for you.