IDLAB, IMEC RESEARCH GROUP AT GHENT UNIVERSITY AND ANTWERP UNIVERSITY
Interoperable Time-Sensitive Networking Towards 6G
Ingrid Moerman & Jeroen Hoebeke
WHY?
2
Demanding Time-Sensitive applications
3
1ms 5ms 10ms 20ms 100ms
AGV/cranes Functional safety Condition monitoring Closed loop control PLC/robot
Live performance Presentation Tour guide Haptics
AV conferencing Entry-level WI Ultimate WI Entry-level SI Ultimate SI AU
D IO/ VI D EO /HAPT
ICS
AR/V R IN DUST
RY 4. 0
3.36Gbps
120Mbps 2.34Gbps
60Mbps 0.150–4.5Mbps / #300
0.15–1Mbps / #10
150-350kbps / #10 – MC #100
E2Elatency inc.proce
ssing
RTT for co mm.
kbit/s - Mbps
Latency /
cycle tim e
kbit/s
kbit/s
kbit/s / < 20μs jitter
< 100kbps
< 1Mbps / 0.2-2ms jitter
Deep immersion
kbps (audio) – Mbps (video)
= motion
Demanding Time-Sensitive applications
4
QoS ≠ CAPACITY
Source: https://www.gsma.com/futurenetworks/wp-content/uploads/2018/04/NS-Final.pdf
Deterministic Networking
[driven by E2E QoS demands]
High Capacity
[driven by technology bounds]
Need for mature vertically integrated radio technology (Main focus on DIGITAL domain)
Development & integration of new radio technology (Main focus on ANALOG domain)
WITHIN CAPACITY BOUNDS RELIABILITY
LATENCY
AVAILABILITY COVERAGE
DETERMINISTIC THROUGHPUT
SPECTRUM EFFICIENCY COST EFFICIENCY ENERGY EFFICIENCY
+
TARGET: verifiable end-to-end, per flow QoS across heterogeneous network domains
§ Verifiable measure performance + take corrective actions ➜ beyond SLA promises
§ Per flow, end-to-end ➜ beyond aggregated QoS per traffic class
Interoperable Time-Sensitive Networking
5
No wireless without wires!
END-TO-END COMMUNICATION
NETWORK INFRASTRUCTRE [composition of network segments]
APP APP
APP APP
APP AP APP
BS
Wired
CN
TN
Cell-free
INFRASTRUCTURE MEDIATED DEVICE-TO-DEVICE COMMUNICATION
ON TIME / IN TIME DELIVERY
SAT AD HOC DEVICE-TO-DEVICE COMMUNICATION
WHAT?
6
End-to-end verifiable Time-Sensitive Networks (TSN)
7
END-TO-END BOUNDED LATENCY
END-TO-END VERIFICATION & AUTONOMOUS MANAGEMENT
Wired Wireless
NET APP
NET APP
< 10ms latency
sub-ms jitter seamless
mobility
MULTIMEDIA NETWORKS
DEMANDING ENTERPRISE NETWORKS
INDUSTRY 4.0 NETWORKS
HOW?
8
OPEN PROTOTYPING PLATFORM
Openwifi
World’s first free WiFi open chip design
Boosting innovation
ieee
Openwifi
10
§ Opensource WiFi chip under AGPLv3
§ https://github.com/open-sdr/openwifi
§ Currently on FPGA based SDR platform
§ Compatible with Linux mac80211 framework
§ SAME as COTS WiFi chip!
§ Unique architecture: SoC, Linux friendly, Fully embedded design
§ Special features beyond COTS
§ CSI, CSI fuzzer, IQ samples…
§ TSN features → THIS TALK
§ Roadmap
§ 802.11ax MIMO, UL & DL OFDMA, 256 QAM,
TWT, From 20 to 40 MHz
TSN Evaluation Kit
11
openwifi Client
TSN switch
User end device User end device
openwifi Client User end device
User end device openwifi AP
Kit available: Q4 2021
SYNCHRONISATION
Time synchronization accuracy
13
Connection openwifi COTS
2.23
µs83.58 ms
2.50
µs82.70 ms
1.78
µs99.41
μsPTP with hardware
timestamping over openwifi PTP with software timestamping over COTS
Parameters No Load TCP Load UDP Load (16.4 Mbps)
Mean (μ) -0.018µs 0.007µs 0.006µs
Standard deviation (σ) 0.840µs 0.824µs 0.919µs 90% percentile (P90) 1.25µs 1.36µs 1.38µs
SINGLE HOP
MULTI-HOP
Validated in w-iLab.t testbed
SCHEDULING
Wireless traffic scheduling
15
A ppl ic at io n
Radio
Netw or k stac k Radio
A ppl ica tio n
Netw or k stac k
Node A Node B
Ra di o
Network stack
AP
1 2 3
4 5
6
Switch 4 HW queues
Dynamic channel access parameters Dynamic traffic flow assignment
1
1 1
2
2
2 3
3 4
4
Time-aware traffic shaping (timed gates)
UL and DL
organization / in-band schedule distribution
5
E2E scheduling
6
4
Wired-Wireless Schedule Cycle Alignment
Flexible airtime slicing
16
Time slot / slice lengths = 128 * 2^TS μs, TS=0..7 from 128 to 16384 μs Cycle = 8192 μs
Slice/Time Slot = 2048 μs
Current cycle lengths = 512 * 2^C μs with C=0..7
from 512 to 65536 μs
Time-aware End-to-End scheduling PoC
17
CO T S d ev ic es
Wi-Fi AP Switch
Wired client
Flow 1 Flo
w 2
Wi-Fi
client 1 Wi-Fi
client 2
E2E latency in scheduled case
18
Time sensitive flow Non-time sensitive flow
End-to-end Traffic scheduling
19
Take AWAY
Cases
Time sensitive data flow E2E latency benchmarking
99t h pe rc ent ile o f E2E la te nc y [ms ]
Openwifi TCP
COTS Openwifi
UDP 20x reduction in 99% E2E latency Wi-Fi AP
Switch
Wired client
Flow 1 Flo
w 2
Wi-Fi
client 1 Wi-Fi
client 2 3,2 ms
END-TO-END VERIFICATION
In-band Network Telemetry (INT)
21
Wireless extensions
Collect node characteristics:
•
Node ID
•
Queue info
•
Processing delay
•
Tx/Rx timestamping
INT source node INT sink node
Wireless link characteristics
•
RSSI, SNR
•
Channel used
•
Retransmission
•
Data rate
•
Back off mechanism elements
End-to-end flow characteristics
•
Flow latency
•
Flow jitter
•
Flow packet loss ratio In-band network telemetry data
Data packet
INT intermediate node
In-band Network Telemetry (INT)
22
Wireless extensions
Collect node characteristics:
•
Node ID
•
Queue info
•
Processing delay
•
Tx/Rx timestamping
INT source node INT sink node
In-band network telemetry data Data packet
INT intermediate node
IPERF 10 Mbps data rate.
INT is added every 5 seconds
Q ue ue fi lli ng
09:28:48 09:30:19 09:31:49 09:33:11 09:34:41 09:36:11 09:37:41 09:39:11 09:40:41 09:42:11 09:43:41 Time of the day
15 20 25 30 35 40 45 50 55
Processing latency [us]
Pr oc es si ng d el ay ( μ s)
In-band Network Telemetry (INT)
23
Wireless extensions
INT source node INT sink node
Wireless link characteristics
•
RSSI, SNR
•
Channel used
•
Retransmission
•
Data rate
•
Back off mechanism elements
In-band network telemetry data Data packet
INT intermediate node
09:28:48 09:30:19 09:31:49 09:33:11 09:34:41 09:36:11 09:37:41 09:39:11 09:40:41 09:42:11 Time of the day
0 1
Retransmissions
0 20 40 60 80 100 120
End-to-end latency [ms]
-46 -43 -40
RSSI [dBm]
FLEXIBLE CONFIGURATION
Relation between INT and TCP congestion window size
25
Maximum queue capacity
Control of TCP congestion based on queue occupation
26
Standard Linux versus modified implementations of CUBIC Algorithm
CWND = Congestion Window size Increased throughput values
Higher bitrate
CHALLENGES & PERSPECTIVES
27
Challenges & Perspectives
§ Time-keeping technology: global time-awareness for all entities & all levels!
§ Rich, expressive and technology-agnostic application- network interface capable to (1) learn fine-grained
application QoS requirements and (2) inform the application about network capabilities
§ Measurability enabling
§ adaptive, fine-grained per-flow, per-hop, E2E monitoring
§ advanced analytics for processing monitoring information (maximal insights with minimum overhead)
§ Autonomous scheduling/configuration techniques across heterogenous network domains capable to cope with
dynamic demands and environments
§ True end-to-end KPIs involving all communication &
application processes (performance & EFFICIENCY)
https://www.cnet.com/tech/computing/facebook-shares-its-time- card-atomic-clock-tech-to-speed-internet-services/
Hardware Software Interfaces Algorithms Standards
…
End-to-end validation Multi-disciplinary research
(co-design)
More info?
SCIENTIFIC PUBLICATIONS
• Xianjun Jiao, Wei Liu, Michael Mehari, Muhammad Aslam, and Ingrid Moerman. 2020.
“Openwifi : A Free and Open-Source IEEE802.11 SDR Implementation on SoC.”
In 2020 IEEE 91st Vehicular Technology Conference (VTC2020-Spring). IEEE.
https://doi.org/10.1109/VTC2020-Spring48590.2020.9128614.
• Muhammad Aslam, Wei Liu, Xianjun Jiao, Jetmir Haxhibeqiri, Jeroen Hoebeke, Ingrid Moerman, E. Municio, P. Isolani, G. Miranda, and J. Marquez-Barja. 2021. “High Precision Time Synchronization on Wi-Fi Based Multi-Hop Network.” InIEEE INFOCOM 2021 - IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS). IEEE.
doi:10.1109/INFOCOMWKSHPS51825.2021.9484531.
• Jetmir Haxhibeqiri, Pedro Heleno Isolani, Johann M. Marquez-Barja, Ingrid Moerman, and Jeroen Hoebeke. 2021. “In-Band Network Monitoring Technique to Support SDN- Based Wireless Networks.”IEEE TRANSACTIONS ON NETWORK AND SERVICE MANAGEMENT18 (1): 627–641. doi:10.1109/TNSM.2020.3044415.
• Jetmir Haxhibeqiri, Amina Seferagic, Ramyashree Venkatesh Bhat, Ingrid Moerman, and Jeroen Hoebeke. 2021. “Tighter application-network interfacing to drive innovation in networked systems.” In Proceedings of the ACM SIGCOMM 2021 Workshop on Network-Application Integration (NAI'21). Association for Computing Machinery, New York, NY, USA, 53–57. DOI:https://doi.org/10.1145/3472727.3472801