January 2012 Winter Meeting Cloud Computing Testbed Session
From Earth Science Information Partners (ESIP)
ESIP Cloud Computing Testbed Initiative, Potomac Room, 4:00 - 5:30, January 5, 2012
Request for Cloud Computing Use Cases
- Peter - moving large amounts of data into the cloud to be processed
- Brian Wilson (JPL) - has a similar need and would like to see moving large data sets into and out of cloud investigated
- operational weather processing - need cloud computing to speed up processing to make it real-time (operational)
Cloud Computing Concept
- 3 options exist
- ESIP established it's own cloud, we rent time on a commercial cloud, government agencies may be willing to offer partnerships to use their clouds
- the final decision might be some combination of these three
- (James) Microsoft Azure is a viable option and they seem willing to collaborate. however, the Windows requirement might deter/limit some applications
- an informal crowd poll showed the need for linux specific applications
- DOE involved in Wyoming supercomputer center and could be a possible government partner
- it was suggested that ESIP pursue an Earth science-cloud challenge, cloud capabilities that are currently lacking in commercial clouds could be pursued through an open technology challenge to ESIP members and industry
- consensus was that all options should be considered in future sessions
- What size is reasonable for a test-bed cloud?
- Brian asked for clarification on what "test-bed" meant - is the goal to create a usable cloud, test cloud features/capabilities, or test applications in a cloud environment before moving to a production-level cloud
- Phil mentioned that we should remember that this is a test-bed and not an operational cloud
- it was generally agreed that the test-bed was a place for ESIP members to test applications in a cloud environment, the end result of any member evaluation should would be documentation of what worked and what didn't
- members using the test-bed should provide ESIP with all relevant information and images if applicable
- it was agreed that initial tests should involve - 1.) data intensive application, 2.) i/o intensive application, and 3.) an application that combines 1. and 2.
