Presentation is loading. Please wait.

Presentation is loading. Please wait.

Cloud Futures 2011 Christopher Alme, Christopher Nunu Dennis Qian, Stanley Roberts Stephen Wong.

Similar presentations


Presentation on theme: "Cloud Futures 2011 Christopher Alme, Christopher Nunu Dennis Qian, Stanley Roberts Stephen Wong."— Presentation transcript:

1 Cloud Futures 2011 Christopher Alme, Christopher Nunu Dennis Qian, Stanley Roberts Stephen Wong

2 Uncharted Skies: Streaming Cloud Real time data streams ground  cloud Process data in the cloud Stream the results back to the ground Huge parallel processing capability Elasticity  dynamic configurability! Cloud Futures 2010

3 Practical Uses Air/Auto Traffic Control Environmental Sensors Inventory Tracking Surveillance Home Automation

4 Initial Project Motivations Interactive Art Dynamically changing inputs, incl. game controllers Scalable number of users Multiple, simultaneous results visualizations

5 The Team 4 students in semester-long software- engineering project No initial experience in C#,.NET, Azure, enterprise systems, etc. Self-organized and self-directed team created and assigned own tasks. Discovery process: Make mistakes and learn from them.

6 User Experience: The Ground Connect Streams to the Cloud Manage Processing of Streams Visual Representation of Results

7 Architecture: The Ground Control Panel Cloud Settings Options Status Stream Mgr Stream Stream Up Stream Mgr Stream Stream Up Stream Mgr Stream Stream Up Display Visualization of Cloud Output Stream Down Visual Representation of Cloud Output Display Visualization of Cloud Output Stream Down Stream Manager Provides a way for the user to connect streams up into the cloud Arbitrary number of running instances Dynamically add/remove Streams User Friendly Stream Manager Provides a way for the user to connect streams up into the cloud Arbitrary number of running instances Dynamically add/remove Streams User Friendly Display Receives streaming output from the cloud Renders the output from the cloud Displays the results on a screen Arbitrary physical location Display Receives streaming output from the cloud Renders the output from the cloud Displays the results on a screen Arbitrary physical location Control Panel Administration of cloud functionality How input streams are used in processing Coordinate the creation of processing graph Assigns display outputs to processed data. Control Panel Administration of cloud functionality How input streams are used in processing Coordinate the creation of processing graph Assigns display outputs to processed data.

8 Message Architecture – The Cloud Feature Extractor Stream Feature Extractor Stream Feature Extractor Stream Mixer Stream Display Adapter Stream Display Adapter 2-way Message Control Process Stream Entry Processing of Streams Stream Exit State Repository Control Process Creates the processing graph structure Provides user access to the cloud structure Manages fault tolerance Control Process Creates the processing graph structure Provides user access to the cloud structure Manages fault tolerance Feature Extractors Stream entry into the cloud Single Stream Number Crunching Converts streams into internal events Forwards events to proper Mixer Feature Extractors Stream entry into the cloud Single Stream Number Crunching Converts streams into internal events Forwards events to proper Mixer Mixers Given an arbitrary lambda Syncs and Processes Streams Linked to produce a graph of operations Mixers Given an arbitrary lambda Syncs and Processes Streams Linked to produce a graph of operations Display Adapters Stream exit point, connection to Display Converts the processing results to Visualization commands Different adapters can be set to produce different results Display Adapters Stream exit point, connection to Display Converts the processing results to Visualization commands Different adapters can be set to produce different results State Repository Shared database Stores processing graph configuration data Allows for re-creation of lost module/connection State Repository Shared database Stores processing graph configuration data Allows for re-creation of lost module/connection

9 Demo Structure Stock value Level Gesture & Button Detect Buy Sell Display Out Filter

10 Real-time challenges in the Cloud Reconstruction of lost modules Modules are not identical! TCP timeout Azure does not assume persistent connections! Lost Connections Azure load balancer can’t make specific connection. Dev Fabric Isolation Doesn’t allow outside connections

11 Questions?

12 Into the Blue was supported by Thank-you for enabling us to take flight!


Download ppt "Cloud Futures 2011 Christopher Alme, Christopher Nunu Dennis Qian, Stanley Roberts Stephen Wong."

Similar presentations


Ads by Google