What's blocking you from doing data science?
Recently, the data science team at an established oil and gas property management company needed to process terabytes of raster images using a MATLAB model. The only problem was that this process would to take more than 30 days due to hardware and licensing constraints commonly associated with a MATLAB cluster. Not only did this expensive hardware require frequent and time consuming upgrades, but usage was a major variable - some months would see constant usage while others had minimal or no usage. The data science team needed a better way to complete their important work, and Pariveda was up to the challenge.
Reduced time to value
The team at Pariveda was tasked with creating a solution that would host the cluster in the cloud using a high-performance computing series of virtual machines, each of which were configured to completely shut down when not in use. The cluster was redesigned from the ground up to run compiled models on as many as 1,000 different worker nodes at the same time, which significantly reduced raster processing time from over a month to just a few short days. Task output was also meticulously logged to storage, enabling a comprehensive analysis of task performance.
Microsoft Azure enables deep insights to improvement opportunities
The significant reduction of raster processing time enabled a shorter feedback loop with geologists, allowing the company to execute investment opportunities much faster and more efficiently than ever before. At the same time, they enjoyed a dramatic cost reduction as the company now only pays for actual compute time used. Additionally, the increased level of insight and visibility into task performance enabled the data science team to better identify available improvement opportunities for the MATLAB model.
Along the same lines, the Azure high-performance computing series virtual machines - along with the Azure Batch - significantly reduced the time it took for raster processing using MATLAB models. Azure Table Storage enabled a deep analysis of task performance, which allowed the company to better identify opportunities to fine tune the model moving forward. It set the stage for continuous improvements across the board.