Deploying OSGi on an Army of CubieTrucksSendrato powerpoint
•Download as PPT, PDF•
1 like•2,576 views
Our challenge was to track the location 200k people at a large festival using custom made wristbands, on a tight budget and preferably with only open source software. In this presentation we will briefly outline the overall system architecture, and subsequently focus on the software stack used to make this work. In essence, we need the versatility of cloud computing, but in a "camping-style data centre" We will show how Apache Felix, an OSGi runtime, provides the necessary mix of dynamism and efficiency to pull this off.
Recommended
More Related Content
What's hot
What's hot (20)
Similar to Deploying OSGi on an Army of CubieTrucksSendrato powerpoint
Similar to Deploying OSGi on an Army of CubieTrucksSendrato powerpoint (20)
More from Frank Lyaruu
More from Frank Lyaruu (8)
Recently uploaded
Recently uploaded (20)
Deploying OSGi on an Army of CubieTrucksSendrato powerpoint
- 1. Wearables at Scale Deploying OSGi on an army of Cubie Trucks Frank Lyaruu
- 2. About me Frank Lyaruu CTO at Senior Technology Architect at @lyaruu 11/17/14 2
- 3. About Dexels • Est. 2001 based in Amsterdam • Enterprise software for large companies • Heavily invested in OSGi • Focused on sports and entertainment 11/17/14 3
- 4. About Dexels • Est. 2013 • Joint venture with Dexels • Custom hardware for large events 11/17/14 4
- 5. • Case study • Hardware stack • Software stack • Development chain • DEMO! 5
- 6. Tomorrowland
- 7. Tomorrowland
- 8. Tomorrowland 2014 Electronic music festival near Antwerp, Belgium • 2 weekends • 8 stages • ~100 hectares • ~100k visitors a weekend • ~ € 250 per ticket 11/17/14 8
- 9. Requirements
- 10. Requirements Design a wristband to do: • Access control • Location • Profiling / feedback • Crowd monitoring • Integrate with 3rd party applications • Something that adds to the UX 11/17/14 10
- 11. Access control • Should always work • Should be secure …simplicity is essential: Passive RFID 11/17/14 11
- 12. Location • Ability to locate individuals • Ability to locate security personnel 11/17/14 12
- 13. Profiling / feedback • How do the visitors behave? • What acts are popular? 11/17/14 13
- 14. Crowd monitoring • Find hotspots in the flow of people • Dashboard for security personnel 11/17/14 14
- 15. UX • UX is key • Comfortable • Match the festival • Suitable for non-nerds 11/17/14 15
- 16. The plan
- 17. The setup • Set up base stations around the terrain • Wristbands will report status • The base stations triangulate and consolidate the data • Visualize and push the data to other applications • Use regular passive RFID for access 11/17c/14ontrol 17
- 18. Something cool • LED lights on the wristbands, integrated with the light show • A button to send Facebook friend requests 11/17/14 18
- 19. The protocol • Time slotted protocol • All base stations synchronise with each other • Visitor bracelets send every few minutes • Staff bracelets send every few seconds 11/17/14 19
- 20. Hardware
- 21. Hardware • One 3V battery for 3 days • Leather • Branded for the festival • < $10 per bracelet 11/17/14 21
- 22. Base station hardware 22
- 23. Base station 23
- 24. § 24
- 26. 11/17/14 26
- 27. Base stations • Heat • Rain • Unreliable power • Unreliable network • Unstable mount points • Electromagnetic interference 11/17/14 27
- 28. Base station deployment 28
- 29. Base stations 11/17/14 29
- 30. Sendrato Infrastructure LLooccaal l“ “ddaattaacceennttrree”” HHaazzeelclcaasstt c clulusstteerr 11/17/14 33rrdd p paarrttyy 30 CCuubbieieTTrruucckk CCuubbieieTTrruucckk CCuubbieieTTrruucckk CClolouudd SSaaltlt M Maasstteerr LLoogg s seerrvveerr MMoonnggooDDbb MMoonnggooDDbb SSuuppeerr c coooorrddininaattoorr MMaannaaggeemmeenntt U UII
- 31. Base station software stack 11/17/14 31 AAppaacchhee F Feelilxix JJaavvaa FFeeddoorraa 2 200 SSaalltt aaggeenntt SSeennddrraattoo A Apppplilcicaattioionn
- 32. Basestation software BBaassee s sttaattioionn FFilitlteerr e evveenntt PPuubblilsishh i nin c clulusstteerr 32 WWrrisisttbbaannddss o Radi RRaaddioio D Drriviveerr OOSSGGi ie evveenntt LLoogg l oloccaalllyly?? Radi o
- 33. Logging • Custom Logback appender to Redis queue • ElasticSearch analytics • Kibana web interface 33
- 34. Salt Stack • Configuration manager • Remote execution • Agent based • Versatile & Fast • Master running in the cloud 34
- 36. Development cycle 36 • Pretty crude but effective: EEcclilpipssee MMaavveenn b buuilidld GGitit SSaaltlt FFilieleInInssttaallll
- 37. Future work • Distributed OSGi • More mature deployment model • Apache Karaf + Cellar • Pluggable network / clustering • Proper 3rd party API 37
- 38. Demo
- 40. Thank you! @Sendrato sendrato.com info@sendrato.com 40
Editor's Notes
- Last name for twitter handle! 1m
- Means ‘wireless’ in Esperanto. (We think) 3m
- Ask questions when you have them
- 5m
- Anyone heard of Tomorrowland? Big deal
- … pretty scary launch customer 6m
- Not the typical enterprise requirements
- People can get hurt when access control fails 9m
- … typical big data questions
- 13m
- Cubie = power pi, with onboard storage ARM architecture
- 17m
- +5
- +8 Also dmx 25 min
- 27m
- 30m
- Server in the cloud because of possible network config problems
- 34m master in the cloud because of possible routing problems agent based faster and more resilient to network probe (Ansible)
- We knew we’d be hacking all weekend
- +3 37 min 41 min
- 42 min.