IoT Silicon Valley - Cityzen Sciences and Cityzen Data presentation
5 likes1,618 views
Cityzen Data is a French startup that focuses on integrating advanced conductive fabric technologies and flexible micro sensors for mass production, with an emphasis on smart textiles like the D-Shirt™ embedded with EKG sensors. They offer an open platform for geo time series data, capable of handling vast amounts of data and supporting applications across various sectors. The company's services include data manipulation, machine learning functions, and a comprehensive ecosystem of third-party applications that prioritize user consent for data usage.
![Example Result
[
{"c":"com.cityzendata.test","l":{".app":"doc.data","origin":"random"},"a":{},"v":
[[1395145993000500,97.5],[1395145993000499,99],[1395145993000498,103.5],
[1395145993000497,104],[1395145993000496,103.5],[1395145993000495,100],
[1395145993000494,99.5],[1395145993000493,101],[1395145993000492,97.5],
[1395145993000491,100],[1395145993000490,99.5],[1395145993000489,100],
[1395145993000488,96.5],[1395145993000487,98],[1395145993000486,98.5],[1395145993000485,94],
[1395145993000484,98.5],[1395145993000483,103],[1395145993000482,100.5],
[1395145993000481,98],[1395145993000480,101.5],[1395145993000479,101]},
{"c":"mozfest.light","l":
{"area":"1",".app":"doc.data","xbeeId":"XBee_40670F0D","moteId":"53"},"a":{},"v":
[[1382714858837000,51.50198796764016,0.005952995270490646,381],
[1382714858830000,51.50198796764016,0.005952995270490646,291],
[1382714828049000,51.50198796764016,0.005952995270490646,230],
[1382714797974000,51.50198796764016,0.005952995270490646,203]}
]](https://image.slidesharecdn.com/20140611-iotsiliconvalley-140612135614-phpapp02/85/IoT-Silicon-Valley-Cityzen-Sciences-and-Cityzen-Data-presentation-38-320.jpg)
IoT Silicon Valley - Cityzen Sciences and Cityzen Data presentation
- 1. Mathias Herberts – CTO Cityzen Data @CityzenSciences @CityzenData @herberts
- 2. We are french startups
- 3. Yes, France has other cities than Paris!
- 4. Opening an international office soon
- 6. Members
- 7. Partners
- 9. Technology Provider Tooling for yarn preparation (gimp)
- 10. Technology Provider Tooling for weaving the conductive yarn in fabric
- 11. Technology Provider Flexible micro sensors
- 12. Technology Provider All at industrial scale, ready for mass production
- 13. Innovation Enabler We work with brands to help them integrate our technologies
- 15. D-Shirt™
- 17. #FFN
- 18. On a flex PCB in phase 1
- 19. More integrated in phase 2
- 20. Less than 50 grams Battery to power the sensors (10 hours battery life) Accelerometers/Gyroscopes GPS Altimeter Data logger BT and BLE connectivity Open platform based on OSGi Gateway
- 22. Companion App
- 24. Platform tailored for sensor data Machines produce massive amounts of data, far more than humans
- 26. A space-time continuum Our platform handles Geo Time Series
- 27. Performance requirements D-Shirt ™ 200,000 data points per hour A platform for creating apps for a large user base, not a select group of analysts
- 28. Performance evaluation TS 101M (10M readings from 1M different sensors) TS 101 (10M readings from a single sensor) 125,000 readings/s ingestion rate per core in the worst case 200,000 readings/s retrieval rate 500,000 readings/s ingestion rate per core 500,000 readings/s retrieval rate
- 30. a language dedicated to manipulating Geo Time Series Data manipulation environment
- 31. Where do we fit in?
- 34. Complete control on data usage
- 35. Ecosystem of third party applications Apps must have user consent - data producer advantage
- 36. // Stack based language // RPL like language (language of HP RPN calculators) // Script results are returned as a JSON array of stack levels // Has macros, conditionals, loops // Has close to 250 functions, both simple and complex ones // Includes data manipulation frameworks // Scripts execute server side, where the data resides
- 37. Geo Time Series have a class and a set of valued labels Data model class{label0=v0,label1=v1} Readings have one of four types LONG DOUBLE BOOLEAN STRING
- 39. MAP – Do computations on a sliding window BUCKETIZE – Normalize the ticks of series REDUCE – Do computations on series partitioned by labels APPLY – Apply n-ary functions to partitions of n lists of series FILTER – Filter series previously partitioned by labels Einstein Frameworks
- 40. POST /api/v0/update HTTP/1.0 Host: api.cityzendata.net X-CityzenData-Token: <TOKEN> <TIMESTAMP>/<LAT>:<LON>/<ELEV> <CLASS>{<LABELS} <VALUE> ... 1400000000000000/37.7833:-122.4167/90000 temp{th=1} 51 1400000060000000/37.7833:-122.4167/90000 temp{th=1} 51 1400000120000000/37.7833:-122.4167/90000 temp{th=1} 52 1400000180000000/37.7833:-122.4167/90000 temp{th=1} 53 1400000240000000/37.7833:-122.4167/90000 temp{th=1} 51 ... 1400244098138164// 心拍数 {id=1} 85 Data ingestion using HTTP
- 41. Exponential Smoothing // Retrieve data 'TOKEN' '~linux.*loadavg.1' // Regexp for selecting class 'host' '10.0.0.1' 2 ->MAP // Label selector '2014-01-01T00:00:00.000Z' '2014-05-01T00:00:00.000Z' 5 ->LIST FETCH // Apply Single Exponential Smoothing 0.0025 SINGLEEXPONENTIALSMOOTHING // Alpha=0.0025 // Only retain 1000 values bucketizer.mean 0 0 1000 5 ->LIST BUCKETIZE
- 43. Pattern Detection // Retrieve data ... FETCH // Detect pattern (pattern extracted via PATTERNS) '......2z' 1 ->LIST 160 // Sliding window 8 // Pattern length 64 // Quantization scale PATTERNDETECTION
- 45. Discord detection // Retrieve some data ... FETCH 64 // Size of discord detection window 16 // Pattern length 512 // Quantization scale 1 // We want to detect a single discord false // Discords are not allowed to overlap (N/A here) 2.0 // Distance ratio between nearest neighbors DISCORDS
- 49. Ability to use Einstein on very large numbers of series Hadoop integration
- 50. Publish/Subscribe mechanism based on WebSockets Plasma Real Time Dashboards Complex Event Processing Storm Esper
- 51. R integration Machine learning functions (anomaly detection and forecasting) Extensions to Einstein WebComponents for dataviz ... Roadmap