1. How we connected a golf grip to the cloud
6/4/2018 5:27 PM
THR2163
How we connected a golf grip to the cloud
Davide Mauri
Director of SW Sensoria
© Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

2. Who Am I? Microsoft Data Platform MVP since 2006
Working with SQL Server from 6.5, on BI from 2003
Specialized in Data Solution Architecture, Database Design, Performance Tuning, High-Performance Data Warehousing, BI, Big Data
Very strong developer background
Loves this community!
President of UGISS (Italian SQL Server UG) for 11 Years
AppDev VG Leader since 2017
Regular SQL Server events
Director SW & Sensoria:
Blog:

3. What Do We Do? Sensors Injected into IoT Garments Wearable Solutions
Hardware/Firmware: Sensoria CORE
Software/Platform/SDK/Cloud

4. Sensoria in the news

5. Why Are We Doing This? IoT for the Humans: “Internet Of Me”
Quantify human activity
Improve human performance
Reduce risk of injury
Golf is one of the most data-rich sport
*everything counts* to make a good shot
Angle, Speed, Impact, Pressure…
And much more!

6. Instrumenting a Golf Grip
It’s one of the most important thing to learn
Measure how pressure is exerted on the grip is very interesting
And even more try correlate the shot outcome with the recorded pressure pattern

7. Instrumenting a Golf Grip
Sensoria Core + 8 Pressure Sensors
Sensoria Core
Pressure Sensor

8. Instrumenting a Golf Grip
Sensors stream at 64Hz
Each sensor has a specific calibration
Data is sent to a receiver app
via Bluetooth 4.0 LE
Application shows real-time data

9. On-Premises: Data Collection
Windows UWP Application
Get data streamed from Sensoria Core
Shows it in real time
Pack it so that they can be send to Azure every second
32Hz/64Hz data cannot be streamed “as is”. Costs will be to high
Data is sent real-time (if network permits) to Azure
Must handle intermittent connection to Bluetooth and Internet. GUI must stay responsive

10. On-Premises: Data Collection
Via the Sensoria SDK Sensoria Core sends streaming data
Data is received in a dedicated thread
Last received value is used to update GUI
Received data is stored in concurrent queues
ConcurrentQueue<T> Class

11. On-Premises: Data Delivery
Two queues help to retain all gathered data and to handle slow internet connections
Every second 64 sensors values are removed from the Receive Queue, packed into one JSON document and put in the Send Queue
Every second a packed document is taken from the Send Queue and sent to Azure

12. On-Premises: Data Delivery
Sensoria Core Data
8 Sensors
2 bytes each
64 samplings per second
+ additional data (Golf Club, DateTime, etc.)
Received
Data
queue
Send
Data
queue
RealTime
Every Sec
Every Sec
Azure

13. Solution Architecture
Lambda Architecture

14. Cloud: Ingestion Event Hub takes care of data ingestion
IoT could be used if you also need to send data back to the device
Or if you need MQTT support
Useful resource to decide which one to use:

15. Cloud: Stream Processing
Processing is done using Stream Analytics
Data is split in four streams with slightly different payload (in order to minimize size, by avoiding duplication)
START, SWING, END
EVAL
EVAL contains evaluation data
Evaluation is done manually by the golfer/caddie

16. Cloud: Serving Data is saved to a Azure SQL Database
A PowerBI dashboard has been created to analyze all session at once

17. Cloud: Serving
Data is also available to the UWP app to “Replay” the shoot

18. Cloud: Machine Learning
Work in progress: analyze sensors data to check if a “best grip” pattern may exist
Include additional external data to improve algorithm
Goal: Give the golfer real-time feedback on its grip (via earbuds)
VERY useful for training
Performance improvement feedback in real-time on the golf course
Gamification

19. Why Azure SQL: BEST VALUE!
For the POC SQL Azure has the best feature/costs ratio, delivering great performance (sort of “battery included” solution)
In-Memory features are perfect for fast data ingestion
Columnstore is perfect for analytics
All data was there so no need to add additional complexity to orchestrate data movements between different azure services

20. Future Options
Other options evaluated and taken into consideration for production phase:
Apache Storm
Handle complex streams
DocumentDB / CosmosDB
quickly save raw JSON data, no matter the shape
Azure SQL DW
easy and faster usage of consolidated and cleaned data
Historical analysis/comparison
Raw Data stored in Azure Data Lake
Future analysis of raw data (data science!)

21. Conclusions http://www.wired.co.uk/article/microsoft-data-driven-golf

22. 6/4/2018 5:27 PM
Thanks!
© Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

23. Please evaluate this session
Tech Ready 15
6/4/2018
Please evaluate this session
From your
Please expand notes window at bottom of slide and read. Then Delete this text box.
PC or tablet: visit MyIgnite
Phone: download and use the Microsoft Ignite mobile app
Your input is important!
© 2012 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

24. 6/4/2018 5:27 PM
© Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.