Always-On Web of Things Infrastructure Dynamic Software Updating
1 like560 views
The document discusses the infrastructure for dynamic software updating (DSU) within the context of the Web of Things, focusing on ensuring low-disruption and timely updates for software applications. It outlines the processes for preparing and executing updates, along with a case study on the Moquette MQTT message broker, highlighting programming efforts and performance metrics. Key technologies and standards, like MQTT and CoAP, are also mentioned in relation to IoT interoperability and updating systems.
Always-On Web of Things Infrastructure Dynamic Software Updating
- 1. KIT – University of the State of Baden-Wuerttemberg and National Research Center of the Helmholtz Association PERVASIVE COMPUTING SYSTEMS, INSTITUTE FOR TELEMATICS, FACULTY OF INFORMATICS www.kit.edu Always-On Web of Things Infrastructure Dynamic Software Updating 2016-11-07, Web of Things Workshop Martin Alexander Neumann, Christoph Tobias Bach, Andrei Miclaus, Till Riedel, Michael Beigl
- 2. 2 16-11-08 e.g. production control, health monitoring Simplify IoT interoperability Common technologies/ web standards HTTP, REST JSON, JSON-LD WebSockets, CoAP, MQTT Martin Alexander Neumann, KIT Web of Things http://forklog.net/the-way-iot-and-blockchain-change-the-world/
- 3. 3 16-11-08 Updating the Infrastructure Cloud services and gateways Highly available, real-time communications Maintenance windows, or rolling updates Software updates Low-disruptive rollout of updates and patches But, especially: timely rollout of security patches Martin Alexander Neumann, KIT
- 4. 5 16-11-08 Ad Hoc Approach enabling … Every update to every installation of a software has to be planned Takes time and effort General At least – approach fits an update to all installations of a SW Better – approach fits all updates to all installations of a SW Dynamic Software Updating Martin Alexander Neumann, KIT Low-Disruptive and Timely Hot/Live Updates Least preferrable with common technologies widely adopted
- 5. 6 16-11-08 1. Preparation Vendor tailors program to DSU system 2. Installation/Rollout Vendor provides a dynamic update (instead of a regular one) to update any old program instances Martin Alexander Neumann, KIT Goal of Dynamic Software Updating
- 6. 7 16-11-08 Conventional Update Martin Alexander Neumann, KIT 𝑣0 update stop restart persist data offline 𝑣1 reload data online disruption
- 7. 8 16-11-08 Transform state in memory → efficient Stop 𝑣0 immediately → timely Transform anything → flexible Use with any software → generic Dynamic Software Updating Martin Alexander Neumann, KIT 𝑣0 DSU offline 𝑣1 online disruption
- 8. 9 16-11-08 Close classes world and load 𝑣1 Compare 𝑣0 to 𝑣1 for automatic transformation Load update code 𝑣0 → 𝑣1 that finalizes transform. 𝑢𝑛𝑙𝑜𝑎𝑑 𝑣0 𝑙𝑜𝑎𝑑 𝑣1 𝑐𝑜𝑚𝑝𝑎𝑟𝑒 Lusagent DSU Phases – Preparation & Cleanup Martin Alexander Neumann, KIT 𝑣0 DSU offline 𝑣1 online prepare cleanup update
- 9. 10 16-11-08 Stop all threads, transform state, release threads Properties or onetime instrumentation for safe state Explore entire heap to find and transform objects Lusagent DSU Phases – Update Martin Alexander Neumann, KIT 𝑣1𝑣0 offline online prepare cleanup update 𝑢𝑝𝑑𝑎𝑡𝑒 𝑐𝑜𝑑𝑒 𝑎𝑢𝑡𝑜. 𝑡𝑟𝑎𝑛𝑠𝑓𝑜𝑟𝑚.
- 10. 12 16-11-08 Develop (one-time) program instrumentation Definition of update points for timely updates Lusagent DSU Workflow (1) Martin Alexander Neumann, KIT Update Points 𝑣0 𝑣1 𝑣0 Update Points 𝑣1 IDE
- 11. 13 16-11-08 Develop update-specific update code/transformations Understand automatically covered transformations Program finalization of transformations Lusagent DSU Workflow (2) Martin Alexander Neumann, KIT 𝑣0 𝑣1 Analysis Stubs for Update Code 𝑣0 → 𝑣1 Automatic Transformation IDE Update Code 𝑣0 → 𝑣1
- 12. 14 16-11-08 Lusagent DSU Workflow (3) Martin Alexander Neumann, KIT Run instance of 𝑣0 Install dynamic update to 𝑣1 (on instance of 𝑣0) DSU System Update Points 𝑣1 DSU System Update Points 𝑣0 Update Points 𝑣1 Update Code 𝑣0 → 𝑣1 DSU System Update Points 𝑣0 1 32
- 13. 15 16-11-08 Case Study – Message Broker Martin Alexander Neumann, KIT Programming effort? Broker performance on update?
- 14. 16 16-11-08 Features MQTT 3.1, QoS 0, 1, 2 Persistent sessions Event-processor, netty.io IoT Framework MQTT adapter Fraunhofer’s SensorThingsServer (OGC SensorThings API), … Releases Martin Alexander Neumann, KIT Moquette – Java MQTT Message Broker Users Version Date SLoC Classes v0.7.0 7/2015 8468 151 v0.8.0 1/2016 9691 166 v0.8.1 2/2016 9670 163 Updates 1 major & 1 minor release
- 15. 17 16-11-08 Programming Effort – Instrumentation Martin Alexander Neumann, KIT Version Date SLoC Classes v0.7.0 7/2015 8468 151 v0.8.0 1/2016 9691 166 v0.8.1 2/2016 9670 163 Changed classes Added lines Deleted lines 3 19 3 2 11 1 same Add update points (& long-running code interruptable) 4 netty.io event processors 4 update points Timer (background thread) persisting database 𝑣0.7 uses additional async. ring buffer in reception
- 16. 18 16-11-08 Update Point Instrumentation Martin Alexander Neumann, KIT
- 17. 19 16-11-08 Programming Effort – Update Code Martin Alexander Neumann, KIT Version Date SLoC Classes v0.7.0 7/2015 8468 151 v0.8.0 1/2016 9691 166 v0.8.1 2/2016 9670 163 Transformed classes SLOC 16 230 3 28 𝑣0.7 → 𝑣0.8 took about 1 day
- 18. 20 16-11-08 Update Code – Moquette v0.7 to v0.8 Martin Alexander Neumann, KIT Field moved Fields renamed
- 19. 21 16-11-08 Update Code – Moquette v0.8 to v0.8.1 Martin Alexander Neumann, KIT Packages changed, last class renamed Transform how persisted messages are stored
- 20. 22 16-11-08 Throughput Latency Martin Alexander Neumann, KIT Updating Performance Update while no connections or messages are lost
- 21. 23 16-11-08 Summary Web of Things Common technologies Updating the infrastructure Dynamic Software Updating In-memory updates, generic Development workflow Case Study: Moquette MQTT message broker Programming efforts Performance properties Martin Alexander Neumann, KIT