There are few industries more dependent on technology than the oil and gas industry. Reservoir simulation technology is responsible for exploring new resources and exploiting existing ones, while also determining the most favorable and cost-effective ways to tap into



Figure 1. Accurate simulations are essential to ensuring a high recovery. Producing water during the drilling procedure renders the well useless and would cost StatoilHydro hundreds of millions of dollars. (Graphics courtesy of StatoilHydro)
known oil and gas reserves trapped beneath the earth’s surface. Geological conditions below the surface equate to a very complex reservoir. Every oil or gas reservoir consists of several sand or carbonate layers (or even channel systems) with varying flow properties. A high number of simulations are needed to determine where to drill wells to ensure a high recovery with as few wells as possible and to minimize the amount of water produced. The stakes of these simulations are very high, since each well drilled costs hundreds of millions of dollars!

Reservoir engineers in the Sub-Surface Division of StatoilHydro ASA, a Norway-based oil and gas company and one of the world’s largest crude oil traders, use sophisticated 3-D simulation applications to search for potential oil-bearing structures in the Earth’s crust by simulating the flow of fluids in oil and gas reservoirs. These applications involve vast amounts of data, large numbers of complex calculations, and require thousands of iterations to produce accurate results. It takes plenty of computing horsepower to meet the needs of these modeling applications. The rising cost of drilling and the high cost of an error in choosing a drilling location put StatoilHydro engineers under increasing pressure to produce ever-more accurate results with their modeling efforts.

The problem
Local computing clusters at each of StatoilHydro’s four Norway locations were not meeting engineers’ needs for increased ease of use in job submission and for additional processing power to run more iterations of their models. Initially, each of StatoilHydro’s locations independently operated its own local computing infrastructure with a set of workstations and server clusters ranging in size from 64 to 400 CPUs. However, the complexity resulting from the number of machines was causing a number of problems. Because each site was investing in its own computing equipment, the clusters needed to be oversized to accommodate peak periods of activity. Furthermore, homegrown interfaces based on old scripting technologies were being used for job submissions and integration with the ECLIPSE application and Platform’s LSF software being used to control the clusters. The scripts were old, difficult to maintain and didn’t provide adequate visibility into the job execution activity going on within the clusters. Finally, the disparity in the relative number of CPUs available to each location caused problems in terms of differences in engineering process consistency, performance and reservoir simulation accuracy.

The solution
Over a 20-day period, Platform Computing implemented a single, division wide computing grid within the Sub-Surface division of StatoilHydro ASA.

First, the Platform LSF software managing each of the four local clusters was upgraded to the latest level to simplify ongoing software management. Then, Platform LSF MultiCluster
Figure 2. Mongstad is one of StatoilHydro’s four Norway locations to now have access to a single, division-wide computing grid as well as the simulation’s group full complement of computers. The grid not only allows computing resources to be managed centrally but also allows engineers to have access to significant compute power anywhere in the world.
software was used to tie the four local server clusters into a single computing grid, giving each engineer user, irrespective of his or her location, access to the division’s entire complement of computing resources, which was fast approaching 1,000 CPUs. Platform worked with Schlumberger’s software development team to integrate ECLIPSE, Platform LSF and EnginFrame, allowing ECLIPSE users to perform significantly more iterations for multi-realization simulations on cost-effective server clusters. The ability to run iterations yields far more accurate results and lessens the possibility of bringing up water rather than oil — even a small error in location could cost millions of dollars.

Platform LSF allows StatoilHydro to dynamically grow the size of its grid by harvesting unused CPU cycles as needed from any of the 185 user workstations across the division that might not be running at full capacity. This accounts for nearly 25% of StatoilHydro’s overall computing capacity and is totally transparent to the workstation users.

Next, a Web portal was implemented based on EnginFrame from NICE, giving users a modern, intuitive interface that greatly simplified the submission of jobs to the grid, provided better visibility into the job execution process and saved time for StatoilHydro IT staff by eliminating the need to maintain their legacy job-submission system. EnginFrame’s ability to support the submission and monitoring of jobs by any skill level of user from anywhere in the world is a key component of StatoilHydro’s grid solution.

As a result of providing its users with access to greater computing power, StatoilHydro is seeing a significant increase in the amount of simulation work taking place throughout the grid — without having to add more computing hardware. Harstad, for example, a StatoilHydro location that would have needed a vast amount of new equipment to support its increasing use of the ECLIPSE application, can now run simulations any time, eliminating the need for additional equipment. When StatoilHydro opens new locations, as they have recently done in Houston, Platform LSF with MultiCluster makes it much easier to implement additional clusters and add them to the grid.

The new division-wide computing grid is essential as StatoilHydro begins to upgrade its cluster operating system from RedHat3 to RedHat4. The system gives StatoilHydro the flexibility it needs, allowing the IT department to take out half its nodes at a time to complete the upgrade while still allowing for the completion of compute jobs.

Previously, local computing clusters at each StatoilHydro location needed to be oversized, and still only gave engineers access to a limited amount of processing power in peak periods, restricting their ability to generate the highly accurate results needed to support cost-effective drilling operations. By implementing a single, division-wide computing grid, all users, regardless of location, can access the grid and the group’s full complement of computers. An intuitive, Web-based portal simplifies global job submission and provides better visibility into job execution. It is no longer necessary to have computer resources located at all sites, resulting in less fragmentation and better usage. The amount of simulation work that engineers are able to perform has been significantly increased. StatoilHydro was able to reduce its need for infrastructure by 20% while at the same time reducing the overall need for maintenance by one man-year.

Comments