Simulation Framework for Wireless Sensor Networks
- But we already have
- NS3/Omnet++
- Cooja
- Hardware Testbeds (FIT/IOT-Lab) (https://www.iot-lab.info/)
- Testing with realistic RF is important
- Cooja cannot give you realistic RF
- Cooja cannot scale to hundreds of nodes
- NS3/Omnet++ gives realistic RF
- But cannot directly interface with real world stacks
- Protocol Design/performance/efficiency
- Dependency on link layer tech
- Asymmetric/Transient links
- Plugs in real-world protocols stacks with realistic RF
- Real-world protocol stacks (Contiki/RIOT)
- you can plugin your favourite OS/stack easily
- With realistic RF simulation (NS3)
- Scale (thousands of nodes without breaking a sweat)!
- Works for 802.15.4 today but can be extended for
- BLE, 802.11-OCB, 802.11s
---?image=docs/res/whitefield_hld.png&size=contain
Our Experience
- Lets talk practical IoT
- We implemented and tested,
- IETF RFCs, and drafts
- IEEE papers
- Unrealistic data with Cooja
- Few use NS3, Castalia-Omnet++
- Requires more effort
- Does not report data on real world protocol stacks
experience(contd)
- Performance@scale
- example, routing/adhoc issues @ scale
- Design@scale
- example, neighbor mgmt issues
- Most of the plugfests, prototypes are limited to few nodes
- Rarely you see pratical prototypes with 100s/1000s of nodes
Example execution
- configuration
- start whitefield
- monitor network
- monitor network with canvas
- OAM - wfshell
- OAM - monitor routing table agnostic of the OS stack used
- pcap, logging
@css[title-top-right](Whitefield demo)