The Design and Implementation of the NDN Protocol Stack for RIOT-OS

Transcription

1 The Design and Implementation of the NDN Protocol Stack for RIOT-OS Wentao Shang, Alex Afanasyev, Lixia Zhang UCLA Presented by Alex Afanasyev December 8, 2016 Workshop on Information Centric Networking Solutions for Real World Applications (ICNSRA)

2 ICN/NDN Edge for IoT Forget about hassle with managing IP addresses Bring IoT semantics to the network layer Name the things and operations on things temperature in the room, humidity on the second floor blood pressure, body temperature max/min/avg ph of soil in specific point of US soil grid Focus on data associated with things, not devices status information or actuation commands Secure data directly W. Shang et. al, "Named Data Networking of Things, in proc. of IoTDI ICNSRA 16, Washington, D.C. 2

3 IoT at the Edge Ultra low cost, longevity constrained battery, low-power networking, limited memory, low CPU SAMR21-PRO: 32-bit ARM, 48 MHz, 32KB RAM, 256KB flash RIOT-OS: multi-platform light-weight OS full C and partial C++ programming environment micro-kernel for multi-threading, priority scheduling, interrupt handling, IPC standard build tools (gcc, make) simulator for testing on Linux PCs gaining a lot of momentum Other platforms Contiki ARMmbed tinyos ICNSRA 16, Washington, D.C. 3

4 NDN-RIOT: NDN For RIOT-OS Optimized for IoT apps Memory efficient packet encoding & decoding Data-centric security support Basic stateful NDN packet forwarding Support for and Ethernet Application API Open source, contributions welcome A few basic examples ICNSRA 16, Washington, D.C. 4

5 NDN-RIOT Architecture Threads NDN app IPC NDN- RIOT module IPC Net Device Driver RIOT-OS Core Sched IPC Interrupt Handler Hardware CPU Timer NIC Peripherals ICNSRA 16, Washington, D.C. 5

6 Memory-Optimized Packet Decoding Shared memory block structure to move packets avoid memory copy in most cases On-demand packet field extraction avoid memory for decoded meta data ICNSRA 16, Washington, D.C. 6

7 Security Support ECDSA micro-ecc library ( secp256r1 curve with 64-byte signatures deterministic signing (RFC 6979) given lack of good entropy on many current devices keys need to be generated outside the device no RSA too much overhead and too expensive to produce signatures HMAC RIOT-OS built-in APIs ICNSRA 16, Washington, D.C. 7

8 Packet Forwarding PIT exact match for interest any prefix match for data (all interests that are prefix of the data) FIB longest prefix match for interest names static compile-time prefix registration IPC-based run-time prefix registration (for local apps) CS any match for interests (a data for which interest is a prefix) compile-time adjustable size (~24KB default settings) FIFO policy Work in progress Extendable / adaptive interest forwarding strategy Support for basic Interest selectors Extend dynamic prefix registration and maintenance ICNSRA 16, Washington, D.C. 8

9 L2 Communication Run directly over layer 2 interfaces IEEE send packets to FF:FF (broadcast) Ethernet (e.g., debugging on native plaform) send packets to FF:FF:FF:FF:FF:FF:FF (broadcast) Simple hop-by-hop fragmentation if needed M SEQ Identification ICNSRA 16, Washington, D.C. 9

10 Application API ndn_app_express_interest ndn_app_register_prefix ndn_app_create ndn_app_run ndn_app_destroy ndn_app_put_data ndn_app_schedule NDN-RIOT Thread ICNSRA 16, Washington, D.C. 10

11 3963 dd faces, Code size & s (PIT, FIB, CS) 15.4 ding & decoding and ing verification L IMITATI L IMITATIONS & F UT Currentl Currently the code is only environm environments. The on anext real on a real IoT device. The curr The current implementa routing routing support or FIB An inter An interesting research vide rou vide routing functional on ang, Z. Wang, Y. Yu, g, Yu,, J.Y. Burke, B. Zhang,. Zhang, Data Networking of rking of EEEConInternational Cononal Design and Implementaplementaaper). OURCE C S OURCE CS ODE ICNSRA 16, Washington, D.C. Theofsource c The source code this wo

12 Basic Evaluations of the Stack Memory overhead on several different platforms for a simple client/producer app Processing speed Processing overhead ICNSRA 16, Washington, D.C. 12

13 Memory Requirements for Functions ndn_name_from_uri ndn_name_append ndn_name_get_size_from_block ndn_name_get_component_from_block ndn_interest_create ndn_interest_get_name ndn_data_create ndn_data_get_name ndn_data_get_content ndn_data_verify_signature ndn_app_run ndn_app_schedule ndn_app_express_interest ndn_app_register_prefix ndn_app_put_data ARMv7-M ARMv6-M ICNSRA 16, Washington, D.C. 13

14 Overall Memory Requirements 40,000 35,000 35,492 35,404 34,092 33,992 30,000 25,000 20,000 15,000 10,000 11,400 11,400 11,400 11,400 5,000 0 Consumer Producer Flash (ARMv6-M) Flash (ARMv7-M) RAM (ARMv6-M) RAM (ARMv7-M) ICNSRA 16, Washington, D.C. 14

15 Stack Performance Numbers Get Data Content 4 µs 4 µs Get Data Name 2 µs 3 µs Verify Data (ECDSA) 294,225 µs 500,115 µs Create Data (ECDSA) 269,314 µs 451,215 µs Create/Verify Data (HMAC) 1,333 µs 1,806 µs Get Interest Name 2 µs 2 µs Create Interest 23 µs 25 µs Append to Name 29 µs 28 µs Get Name component 7 µs 8 µs Get Name size 11 µs 13 µs URI to Name 282 µs 184 µs 1µs 10µs 100µs 1,000µs 10,000µs 100,000µs 1,000,000µs IoTLab-M3 SAMR21-XPRO ICNSRA 16, Washington, D.C. 15

16 Stack Processing Delay IEEE MTU: 102 bytes Fixed data rate: 250 Kbps Fetching from remote node (generated data) Fetching from remote cache 100 bytes 196 bytes Fetching from local cache 0 ms 50 ms100 ms150 ms200 ms250 ms300 ms350 ms ms ICNSRA 16, Washington, D.C. 16

17 Work in Progress Energy consumption evaluation / optimizations Advanced forwarding strategy supportdata discovery Nearby data discovery Pub-sub API on top of Interest/Data exchange Help welcome! ICNSRA 16, Washington, D.C. 17

18 Use Cases and Other IoT-Related NDN Efforts NDN-BMS: encryption-based access control Wentao Shang, Qiuhan Ding, Alessandro Marianantoni, Jeff Burke, Lixia Zhang. "Securing Building Management Systems Using Named Data Networking. In IEEE Network, Vol. 28, no. 3, May NDN-ACE: authorization framework for actuation apps W. Shang, Y. Yu, T. Liang, B. Zhang, and L. Zhang, NDN-ACE: Access Control for Constrained Environments over Named Data Networking, NDN Project, Tech. Rep. NDN-0036, Revision 1, December NDN-IoT: toolkit for NDN dev on Raspberry Pi NDN on Arduino: minimal app for Arduino ICNSRA 16, Washington, D.C. 18