Written by Charles Meertens
21 December 2018
UNAVCO and IRIS are in the third year of an NSF EarthCube project called GeoSciCloud that is evaluating the potential of hosting large facility-scale services in the commercial cloud environment (Amazon AWS) and the private cloud (NSF XSEDE Jetstream resources at Texas Advanced Computing Center and Indiana University).
Both IRIS and UNAVCO have established on-premises infrastructure to manage, distribute, and curate diverse earth observation data spanning many orders of magnitude in data volume and covering a global scale. Data management includes providing continuous real-time data as well as file downloads from remote sensors and space agencies.
UNAVCO services tested in parallel on cloud-based services include data serving via FTP from a large data store, operation of a metadata database, production scale processing of multiple months of geodetic data (Global Navigation Satellite System and satellite Synthetic Aperture Radar data), web services delivery of quality checked data and products, large-scale compute services for event post-processing, and serving real time data from a network of 800-plus GPS stations. Tests on AWS have been completed and comparisons with XSEDE are underway.
IRIS evaluates cloud services by deploying a set of seismological data and key web service infrastructure to each cloud resource. Web services were deployed that provided bulk access to raw time series data and on-demand processing of data. Particular attention was paid to use cases such as highly concurrent, parallel data access that can be supported by IRIS’s own data center.
This project has identified technical challenges of migrating facility-scale services to the cloud. Additionally, cost benefits, performance, and other metrics are used to assess the effectiveness of cloud services illuminating the potential and limitations of running facility scale production services in the cloud.
This material was presented in more detail at the annual AGU Fall Meeting in Washington, D.C. on Friday, December 14, 2018. The slides from this presentation are available for viewing as a pdf:
Last modified: Friday, 21-Dec-2018 21:14:01 UTC