Leading Energy Company Picks Voltaire to Meet Challenges of Oil & Gas Exploration and Production

While operating in a competitive industry, a leading oil and gas exploration company in Asia needed a clustered solution that would deliver the best performance possible for its upstream applications.  The interconnect fabric would need to be able to support larger compute and storage units to enable deeper and faster analysis.  The sought-after competitive edge was found in a solution delivered by Voltaire and HP, resulting in more efficient production and an exploration process that is now faster than ever.

Customer Challenge

The tightening oil supply and ever-increasing energy prices make the value of rapid discovery and decision-making more important than ever. To ensure the most efficient and accurate exploration possible and the most efficient extraction from existing reservoirs, the customer needed a flexible, high performance computing platform that would enable deeper and faster simulations and provide higher precision visualization capabilities. This would help to keep the company competitive in all its main areas: oil and gas exploration and its development, production and marketing.

Having previously deployed and operated clustered systems that were based on Gigabit Ethernet, past experiences were analyzed and a list of requirements was developed. It was determined that a solution was needed to:

• Get the best performance possible from key upstream applications,
• Provide the bandwidth and compute power required for an advanced visualization system,
• Leverage industry-standard building blocks to maximize price-performance and open standards to provide compatibility with existing applications,
• Support larger compute and storage units to support deeper and faster analysis, and
• Enable a scale-out infrastructure that brings high performance to both interprocessor communication and storage.

A key challenge would be to reduce the complexity of both the large-scale grid installation and its on-going operation.

The Voltaire Solution

After all requirements for the new solution were weighed, Voltaire solutions were selected for a high performance, low latency network for a 150-node HP Cluster Platform based on HP DL145 G3 servers with AMD Opteron dual-core processors.  The Voltaire solution consists of a Grid Director ISR 9288 multi-service switch and 150 Voltaire HCA 400EX 10 Gb/s InfiniBand adapters.  Voltaire multi-service switches offer integrated InfiniBand, GbE and Fibre Channel connectivity, which was an important consideration for scaling out the cluster.  For storage File I/O, the cluster uses HP SFS which delivers a 10X performance boost with Voltaire solutions when compared with GbE. The cluster also uses ScaliMPI 4.2.2 and runs Schlumberger ECLIPSE.

The strong partnership between Voltaire and HP allowed the building blocks of the grid to be assembled, configured and partially cabled at the HP factory as offered by the HP XC solutions. As soon as the server racks were staged at the customer’s data center, it took less than four days for the entire 150-node grid to be fully operational. The speed of the deployment was achieved in part through the use of Voltaire’s GridVision management platform that provided essential information such as cabling verification (matched against the topology design) during the installation phase. During ongoing operation, GridVision detects traffic hot-spots (potential congestions) and reroutes traffic within the switching infrastructure.

Why Voltaire?

Working with HP, Voltaire provided the customer with a complete clustered solution that combines industry-standard compute nodes, director-class switches that leverage state-of-the-art InfiniBand clustering technology, fabric management, host channel adapters, and software.  Voltaire Grid Director switches provided a distinguished advantage because they offer integrated InfiniBand, GbE and Fibre Channel connectivity in a single chassis.  This enables MPI and storage traffic to run on the same network, a capability that is not available with Ethernet or proprietary fabrics.

Voltaire Grid Director switches are also highly flexible and can be configured as fully non-blocking or to have lower backplane capabilities depending on application or budgetary requirements.  Grid Director switches also can be upgraded to fully non-blocking without disruptions and also feature cable management solutions – an important consideration for the 150 node cluster.

By enabling IPC and high performance storage on a single network, the limitations of NFS were abandoned and the customer moved to HP SFS with InfiniBand which enables them to achieve much greater scalability.  All of the 150 nodes can easily write to the same directory (needed for functions such as checkpoint/restart) and get unmatched performance: 350MB/s per client (10X faster than GbE) and 800MB/s on storage nodes (OSS) (also 10X faster than GbE).

The Results

Upon completion, the new grid was quickly provisioned to run upstream applications including Schlumberger ECLIPSE for optimization runs.  In a matter of days after the system was deployed, reservoir simulations were conducted that resulted in higher production efficiency from the deeper calculations performed. Production efficiency increased in average from 60% to 75% compared with the older infrastructure.  

Early experience from the field also shows that the exploration processes have been improved significantly. Analysis is now completed 30% faster than it was previously, reducing the costs of oil exploration dramatically.

In visualization, the customer has experienced tremendous performance gains over their previous cluster in terms of MPI, Storage and Visualization applications.  Now the customer has reached effective resolution in rotation of 800dpi vs. 400dpi on their Ethernet cluster and a rotation rate of 5’/s (Voltaire) vs. 1.8’/s (Ethernet).

Key Results Delivered by Voltaire Grid Director™ Switches:

• Exploration applications run 30% faster
• Oil production efficiency from reservoirs has increased to 75% from 60%
• Upstream applications IT infrastructure is now more scalable than ever – the application can now utilize thousands of CPUs for a particular analysis or simulation
• Unmatched visualization precision - 3D models manipulation is faster and image resolution has increased to HD levels.