Seismology as a Service: Portable Product Generation at the Southern California Seismic Network Using Service-Oriented Architecture and Cloud Computing
Description:
Rapid advancements in machine learning, cloud computing, and new data types such as DAS bring both opportunities and challenges for seismic networks. These algorithms and technologies can enable networks to produce more accurate data products in a more timely, and robust manner. Recently, the Southern California Seismic Network (SCSN) began incorporating machine learning picks from the PhaseNet algorithm to produce higher quality phase arrivals and more accurate origins prior to analyst review. Work presented at the 2023 SSA Annual Meeting showed how cloud native services can make processing scalable in a large sequence. For a network to take advantage of these continually evolving efforts, its staff must be able to evaluate different algorithms, which require efficient deployment and testing. Networks using the same code and algorithms may wish to operate them in a cloud, hybrid, or on-premises environment to fit their individual considerations, such as network bandwidth, computational resources, or budget. To deal with this variability, it is important to make network processing components modular and agnostic to the infrastructure. This would expedite testing and enable faster incorporation of new scientific algorithms, software, and other infrastructure into operational earthquake monitoring.
In this presentation, we configure the AQMS automated origin refinement process, “hypomag”, to a service-oriented architecture. This workflow refines and adds phase picks around expected time windows based on the real-time origin to improve the origin and magnitude. We organize these processing steps into services that could be hosted in the cloud or on premises. We examine modularity by incorporating different picker services that produce picks based on different algorithms. We also examine scalability and cost considerations if these services are deployed in AWS. This work will inform network operators on methods to make their processing flows more adaptable and versatile. It will also familiarize researchers with what is required to make new algorithms a part of an operational environment.
Session: Network Seismology: Recent Developments, Challenges and Lessons Learned - IV
Type: Oral
Date: 5/2/2024
Presentation Time: 10:45 AM (local time)
Presenting Author: Ellen
Student Presenter: No
Invited Presentation:
Authors
Ellen Yu Presenting Author Corresponding Author eyu@caltech.edu California Institute of Technology |
Gabrielle Tepp gtepp@caltech.edu California Institute of Technology |
Ryan Tam rwtam@caltech.edu California Institute of Technology |
Aparna Bhaskaran aparnab@caltech.edu California Institute of Technology |
Shang-Lin Chen schen@caltech.edu California Institute of Technology |
Jimmy Choi jchoi5@caltech.edu Nationwide IT Services |
Allen Husker ahusker@caltech.edu California Institute of Technology |
|
|
Seismology as a Service: Portable Product Generation at the Southern California Seismic Network Using Service-Oriented Architecture and Cloud Computing
Category
Network Seismology: Recent Developments, Challenges and Lessons Learned