Seismology in the Cloud: Prospects and Applications

Session: Applications and Technologies in Large-Scale Seismic Analysis
Type: Oral
Date: 4/28/2020
Time: 08:45 AM
Room: 120 + 130
Description: 

Cloud computing has a promising future for high throughput computing (HTC) in seismology. By storing community seismic data, such as raw signals or curated and labeled data sets (often > 1 TB) in the cloud, researchers can avoid the largest bottleneck in data intensive research: data transfer via the internet from FDSN data centers. We test cloud performance using the Southern California Earthquake Data Center (SCEDC) Open Seismic Dataset on Amazon Web Services (AWS). The SCEDC open data set contains 100+ TBs of continuous ground motion velocity and acceleration seismic waveforms recorded by the Southern California Seismic Network (SCSN) going back to 1999. We achieve data transfer rates on the scale of Gigabytes per second per instance when transferring data from AWS Simple Storage Service (S3) to AWS Elastic Compute Cloud (EC2). Such transfer performance rapidly improves the time-to-science and ability to test algorithms on large data sets. We will also present examples of emerging cloud technology with promises for seismology, including serverless computing, multi-GPU deep learning and an AWS app that interfaces with IRIS-WS and uses Julia to run cross-correlations in the cloud. We believe cloud computing is the best option for decreasing time-to-science and increasing reproducibility in data-driven seismology.

Presenting Author: Tim Clements

Authors