Download presentation
Presentation is loading. Please wait.
Published byGerard Barnett Modified over 8 years ago
1
U.S. Department of the Interior U.S. Geological Survey Dealing with Streaming Data CDI Connected Devices Working Group Jeff Allen and Dell Long USGS Fort Collins Science Center
2
Technology Challenge Lots of data now available in streams Social media Environmental parametrics Observations from many simultaneous sources More variables as part of stream Much higher expectations for technology Deal more sources Provide more analyses Present analyzed results sooner
3
Traditional Workflow Source sends signal to server Signal filtered, preprocessed, and pushed to database for persistence Data post-processed Products generated -- dispatch, display, and distribution Excruciatingly sequential Scale by buying bigger servers, multi-threading Memory allocation always an issue
4
Rethinking Options – Reactive Applications Based on Reactive Manifesto Architecture to build event-driven, scaleable, resilient, and responsive applications Designed to take advantage of cloud environments Works off Actors Lightweight concurrent entities Each has a distinct state and behavior Fail fast, recover quickly, can run remotely
5
Legacy @USGSTED Workflow
6
Reactive Approach Akka – toolkit and runtime Build highly concurrent, distributed, and fault tolerant event-driven applications Small memory footprint; ~2.5 million actors per GB of heap Apache Storm – distributed computation consumes streams of data processes streams in arbitrarily complex ways repartitions streams between computation stages Hadoop analog for streaming data
7
Storm Spouts and Bolts
8
@USGSTED Alternative Workflow
9
Extending USGS/DOI/Federal Capabilities Provide @USGSTED partners with additional capabilities FEMA needs beyond initial dispatch Multiple social media feeds Proximity sensors Observation networks Way to deal with almost any flood of distributed simultaneous signals
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.