4. Introduction
Confidential 4
• Data - massive, it is complex and variable. Billions of sources.
• Machine learning to organize the data and generate insights.
• Learned self-correction and adaptation.
• Reduce energy and optimize asset utilization.
• Edge: Constrained compute platform, Limited connectivity, Mobile or static.
• Send device events
• Receive device command
5. Confidential 5
Edgent
• Open Source @ ASF, initially by IBM.
• programming model and micro-kernel style runtime
• embedded in gateways and small footprint edge devices
• enabling local, real-time, analytics on the continuous streams of data
• Works in conjunction with central analytic system
• Intelligence in data in propagation.
• Connectors
• Development mode – Web console
8. Echo System
Confidential 8
Edge device
Eventinjest
Runtimeanalytics
DeviceEventhub
Send
Events
Receive
Cmds
Applications
Time
series DB
Centralized
Runtime-analytics
server
Reporting
Engine
DB
12. Advantages
Confidential 12
• Multiple Independent applications
• Publish subscribe – topic based
• Provide new Services to client
• Functionality added using system application
• Applications can be registered without being started
• Start the application
• Device commands can control the application
• Send additional data related to problem
• Temporarily reduce resource consumption
13. SDK
Confidential 13
• Micro kernel style – oplet
• 512MB – Demoed on Rasberry PI
• Right now in java but can extnd to Swift, Scilab, Python
Wearable Tech
Connected Home
Industrial Internet
Healthcare
In-store Retail
Connected Car
Data from billions of interactions between devices and people is not only massive, it is complex and variable.
Device commands?
Switch on off certain features
Traffic ahead
Modify the feature to send data every 5min instead every 1min
It can choose its own data sources and decide which patterns and relationships to pay attention to.
Connectors - MQTT, HTTP, Web sockets, JDBC, File, Kafka, IBM Watson
Event injector :
Any imp events
Change in state
Sensed value
Runtime analytics:
Local
Quarks
Device event hub:
Kafka client
Flume
Monitor app – add functionality using system application
Applications can be registered without being started
Device commands can control the app
SilverHook Powerboats, maker of some of the world's fastest monohull watercrafts, wanted to use sensor data collected from racing boats to improve the decision-making abilities and safety of racers and to enhance the fan experience. In powerboat racing, racers rely on telemetry data from their boats to formulate strategy and make safety-related decisions. However the high speeds and the pounding against salt water often taxes on-board equipment and drivers. To mitigate such risks, monitoring data is fed back to analytics engines, which provide real-time alerts such as engine performance issues, potential battery failure or even biometric data such as driver exhaustion.
Sensors on SilverHook's racing boats provide more than 80 sources of data, gather measurements at 100 times per second and then transmit the data to on-board computers at five times per second for on-shore teams. But there wasn't a way to collect, distill and deliver insights in a useful format. In collaboration with IBM and Dataskill, SilverHook Powerboats employed a solution that streamed data to a cloud-based analytics solution. Racers can now have access to real-time information while they race, helping them to make adjustments to on-board equipment while they race. The rich visual interface can also allow fans view boat locations, speed and leaderboards in real-time.
"Quarks represents a natural extension of our streaming analytics project. Quarks can be deployed directly on our boats to perform analytics locally. The result is faster insights, which will ultimately help us win the race," said Nigel Hook, co-founder and CEO of SilverHook Powerboats. "Another benefit is we remove dependence on communications networks, which can be unreliable on the water. Quarks offers analytics at the edge so we can pursue a common streaming analytic model across our boats and our central streaming application."
Reduced Communication Costs
Edgent performs real-time analytics on the edge device, separating the interesting from the mundane, so you don’t have to send every sensor reading over a network. If 99% of readings are normal, Edgent detects the 1% anomalies and just transmits those for further processing.
Local and Faster Time to Action
Edgent makes devices more intelligent, enabling them to take immediate action. For example, a connected vehicle running Edgent can adjust traction control based on the weight of the cargo/passengers.
Learning From Related Devices
Edgent enables connected devices to learn from related devices. For example, a truck maneuvering roads in Oregon can adjust based on the data received from trucks operating under similar loads and conditions in Colorado; data such as altitude, cargo, weather and traffic conditions.