Amazon Web Services Re:Invent 2018 oil and gas track

BP reports re-architected ‘cloud-first’ network. Amazon demos accelerated seismic pattern recognition, ML-backed geosteering and MWD.

Speaking at the 2018 ‘Re:Invent’ Amazon Web Services user group meeting in Las Vegas, Mehdi Far outlined how Amazon’s cloud is ‘transforming’ upstream oil and gas with hosted machine learning. One customer use case compared on-site seismic pattern recognition model training that took 8 hours on a fairly high spec PC with CUDA graphics. The same took a mere 30 minutes on a single EC2 instance. Amazon’s SageMaker ML platform’s ‘out-of-the-box’ templates were used to train and deploy a customized deep learning model for seismic applications. More generally, AWS compute resources are claimed to cut vanilla seismic processing times ‘from months to days’. Far’s salt pick demo ran on a public domain seismic data challenge set by TGS and available on Kaggle. Another application revolves around ingestion and management of logging operations where multi-terabyte data sets are analyzed in real time for MWD and geosteering applications. These leverage AWS storage and ML solutions which allow for millisecond model updates. AWS provides data management and analytics to clients without major in-house resources.

BP’s Paul Schuster and Alaa Nasser with help from AWS’ Paul Burne outlined BP’s ‘quantum leap’ towards a cloud-first network. To support its thousands of remote sites, BP has re-architected its operating model for delivering network services. Achieving high-bandwidth low-latency connectivity between BP and AWS has involved a major revision of security segmentation, access policies, trust boundaries and connectivity to untrusted external networks. BP has developed a modular backbone providing data center-independent and carrier-neutral services proxied centrally. Granular segmentation, malware protection and fine-tuned access policies have led to trust in the cloud. AWS has enabled speedy deployment of a performant cloud. Key learnings include the benefits of distributed services, ‘centralized by exception’ and similarly the use of cloud-native products, ‘augmented by exception’. The merits of a ‘Devops-aligned’ operating and sourcing model and an ‘enterprise-wide and continuous’ architecture were also vaunted. BP is now working to extend its solution with support for cloud federation. Another Amazon presentation covered ML predictive quality management in the downstream. More on the AWS Re:Invent portal.

Click here to comment on this article

Click here to view this article in context on a desktop

© Oil IT Journal - all rights reserved.