Oil and Gas Industries & HPC

The oil industries are divided into Upstream & Downstream activities. Upstream refers to pre-discovery activities & pre-processing activities, while Downstream refers to post-discovery activities & processing activities.

Most computations are done in the upstream activity and can be divided in to 3 main classes.

  1. Field data collection
  2. Data processing
  3. Result dispatch & drilling

The complexities of these activities are described below.

Field data collection

This phase includes sonic detonation and collection of data from either earth bound rigs or sea vessels. Operations moves from place to place, using either compressed air (>1000psi) in the ocean, or detonation on land. Sonar data is collected & sent for processing. The data size are in the order of 10-30TB & they are transferred to key data processing sites for analysis. The data is currently captured on tape.

Data processing

This stage involves 2 parts.

  1. processing of the raw data into earth strata charts
  2. visualization of strata chart for location identification

A) processing of data is a intensive job that can either be run as a high throughput job or MPI job. This depends on the application used. the computation starts typically with identification of independent jobs and then allocating them into runs based on priority. Once identified, the job will be sent to a scheduler & normal HPC rules will apply. Several mathematical algorithms exist to assist the computation and are built into software like PARADIGM. The results obtained are subject to a window and there is some dependence on forward & reverse data. Jobs can be divided into sectors or processed as a whole. Results are based on a std format output, and is in the range of GBs.

Paradigm is made up of 2 software pieces called DISCO & FOCUS. The former is a computation tool where the latter is a X-GUI.

B) Visualization of strata results builds 3D images to help to identify possible drilling locations. This data comes in the form of strata layer flat files that form a “cube”. Different views are built & analysis as an individual or team is sought. Software packages includes LANDMARK. Sharing of visualization across teams becomes an issue. Better visualization & teaming should lead to better guesses for drilling locations.

Result dispatch & drilling

Once the location has been identified, location of drill comes in the form of GPS data & maps to rigs & vessels. Each drill cost typically US$15million. Slower data processing in the previous stages will lead to idle rigs & vessels that contributes to operational cost. Hit ratio ranges from 1:4 -> 1:5

External Presentation

Attached is a external presentation i found in the past on the Petroleum Industry and HPC. Posted here for reference only.

Presentation on HPC and Petroleum industries