Hyperparallel Seismic Processing

Oil and Gas organizations have demanding high-performance applications, both upstream and downstream.  Most of these applications run as high-performance computing systems (HPC) aggregating multiple servers computing power to deliver to solve compute-intensive problems.

As seismic processing applications continually advance their analytic techniques to derive more accurate actionable information, organizations are adding AI and machine learning more accurately analyze, interpret, and make decisions on higher resolution data.

The performance bottleneck for HPC is storage. Seismic applications read and process vast amounts of data concurrently with throughput requirements reaching TB/sec.  SAN storage cannot keep up.  As a result, Oil and Gas HPC systems deploy direct-attached storage (DAS) with software-defined storage (SDS) nodes, or via a clustered parallel file system storage such as IBM® Spectrum Scale™ or open-source technologies like Lustre® or BeeGFS®.

There are several significant issues overlooked with HPC DAS. These include:

  • SKU Sprawl

HPC server node SKUs change frequently.  The server bought last year is different from what’s available today.  New nodes have different CPUs, memory, SCM, NVMe SSDs, SSDs, HDDs, and microcode.  Lots of SKUs over time create excessive numbers of combinations, spares, management complexity, and troubleshooting frustration.  Costs increase.

  • Excessive Data Movement

Poor storage performance density leaves few cycles for data protection.  SDS steals CPU cycles and memory from HPC server nodes. Data protection (snapshots and multi-copy mirroring) are CPU and memory intensive stealing resources from nodes. Moreover, multi-copy mirroring impacts network performance.

  • Minimal Storage Services

To alleviate the DAS performance impacts above, organizations eliminate trusted SAN storage management services such as RAID, snapshots, thin provisioning.

Pavilion took a hard look at seismic processing requirements and realized that seismic data storage needed a fundamental rethink. Oil and Gas organizations demand extremely high throughput storage performance, excellent performance density, high capacity density, and flawless data protection; they also need a low total cost of ownership. This is what Pavilion is now delivering its Hyperparallel Flash Array (HFA) to the Oil and Gas community.

Proven in the most demanding HPC environments with IBM® Spectrum Scale™ and other parallel file systems, Pavilion delivers 20 controller pairs, 90GB/s throughput, 20 million 4K IOPS, (40) 40/100Gbps Ethernet or InfiniBand ports, and 1.1PB of NVMe flash in 72 drive slots in 4RU.  That’s a performance density unmatched by any storage vendor today.  Combine this with RAID-6, SWARM recovery, Zero-Space Snapshots, and Thin Provisioning along with an OpenChoice of NVMe SSDs and Pavilion’s HFA stands alone as the ideal choice for seismic processing, AI and ML in the Oil and Gas industry.

To learn more about Pavilion’s Solutions for Oil and Gas, download this free white paper from Dragon Slayer Consulting.