Experimenting Deterministic Communication Services on the Operator Side

Experimenting Deterministic Communication Services on the Operator Side Dr. Luis M. Contreras, Marta Blanco Caamaño Telefónica CTIO / Transport Dept. DetNet –TSN Workshop Madrid, Spain, 26/07/2025 [email protected] 2 Agenda •Context and motivation •Experiments and work in progress •Interconnection of TSN islands •TSN FRER in a 5G Mobile Packet Core Environment •DetNet PREOF •Basic characterization of FlexE performance •Characterization of critical applications •Path Computation based on Precision Metrics •Conclusions 3 Context and Motivation 4 Context and motivation •The main driver for the evolution of telecommunication networks has been the continuous increment of the offered throughput as main key performance indicator. •Network planning and operation has been traditionally focused on capacity upgrades and bandwidth reservation. •However, a new breed of services has emerged (e.g. VR, AR, industrial, etc) demanding more careful consideration of latency and jitter,as relevant parameters to ensure correct service delivery, which requires to define, measure and enforce relevant network KPIs for those. •Guaranteed delivery is also necessary for some of those use cases where reliability is essential •Bottlenecks in networks will never disappear •However we can mitigate and minimize their effects, or at least keep them under control •So, how to integrate deterministic services in the network of an operator? •The following slides present some of the initiatives explored by Telefónica CTIO in this direction 5 Time-critical communications Source: Ericsson Mobility Report, Nov 2020. 6 Reference of low latency demanding use cases Source: work of MEC initiative at Telefonica CTIO and V2N latency references at 3GPP (TS 22.886). Max** Recomm. Max** Recomm. Min*** Recomm. Min*** Recomm. Holoverse 20 < 10 120 380 160 480 Holoverse 20 < 10 160 480 10 30 Holoverse 100 < 50 3 3 0,05 0,1 Karaoke 45 15 10 3 0,5 0,15 0,128 0,512 Immersive streaming for live events 600 200 15 20 15 20 Augmented Reality 33 510 010 40 310 Drones 80 40 30 10 17 50 60 80 Face recognition: Surveillance 100 50 10 40 10 30 Automated Guided Vehicles (AGVs) 30 < 15 15 20 15 20 Metaverse 15 20 High density Vehicle platooning 10 Vehicle platooning 25 Automotive: 10 25 25 eV2X 5 1 20 Automated driving 25 Automated driving 100 0,5 0,5 Video data sharing for assisted and improved automated driving - human visual system 50 10 10 Video data sharing for assisted and improved automated driving - machinecentric video data analysis (e.g. ultraaccurate position estimation) 10 100 700 100 700 Throughput DL [Mbps] Throughput UL [Mbps] Use case Latency [ms] Jitter [ms] 7 Scope Icons: Copyright © 2020 Telecom Infra Project, Inc. and used with permission. Unauthorized use is prohibited Mast Internet HL1/HL2 Data Center (Service Platform Endpoint) HL4 HL5 eNodeB Optical Transport Access Aggregation Backbone Core (& Edge) Internet OLT Core (MME, HSS, SGW, PGW,…) Service latency and jitter requirements apply to all elements included in e2e service delivery i.e. application, service platforms, network access, transport or mobile core. A holistic approach will be important for an optimal service and network dimensioning in terms of efficiency and performance. Platforms Source: 5G-ACIA, “DetNet-Based Deterministic IP Communication Over a 5G Network for Industrial Applications” Transport network Cloud Compute Party 1 (e.g. User) Capture devices … Replay devices Fixed, mobile, wireless, wireline access Edge Compute … Access network Party 2 (e.g. Visitor) Capture devices … Replay devices Fixed, mobile, wireless, wireline access Edge Compute … Access network Party N Capture devices … Replay devices Fixed, mobile, wireless, wireline access Edge Compute … Access network … User Metaverse-enabler Networking ●Processor power ●Media flow sync delay ●Frame rate ●Bit rate ●Resolution ●Codec compression artefacts ●3D processing ●Display resolution ●Decoding delay ●Async/sync rendering delay ●De-jitter/playback buffer ●Camera sensor noise ●Sensor delay ●Encoding delay ●Transmission queuing delay ●Call setup delay Network ●Loading ●Congestion ●Packet loss ●Jitter ●Radio link signal ●Link speed and variation ●Throttling ●Switching/forwarding ●Propagation delay Edge server ●GPU/CPU processing power ●Workloads ●Cache queuing delay ●Codec encoding parameters Cloud server ●GPU/CPU processing power ●Workloads ●Cache miss delay ●Server queuing delay ●Server processing delay ●Codec encoding parameters Impacts Source: QoE/QoS Measurement Framework & Use Cases QoS Requirements, FYUZZ 2023 9 Sources of delay (and jitter) •Latency for a path in a live network is variable, following a statistical distribution •Multiple sources of delay influence the overall measured latency (and jitter) •Average latency is usually taken as reference value, but it is not sufficient for proper assessment of observable latency as experienced by customers •Another approach is to characterize a set of packet latency samples using order statistics, e.g., minimum (P0), 25th percentile (P25), median (P50), P90, P99, maximum Sub-optimal routes/paths, name resolution, content placement, service architecture, etc. MTU discovery, NAT delay, loss recovery, congestion notification, etc. Structural delays End-points interaction Signal propagation, serialization, delay, switching delay, queueing delay, etc. Path delay Capacity, carrier aggregation, multipath, etc. Link capacity Operating system delay, head-of-line blocking, buffering, etc. End-host Sources of delay Multiple directions to take: from planning to engineering, including introduction of novel data planes and more efficient architecture Source: B. Briscoe et al., "Reducing Internet Latency: A Survey of Techniques and Their Merits" IEEE Comm. Surveys & Tutorials, vol. 18, no. 3, pp. 2149-2196, Third Quarter 2016 16 Work in progress TSN FRER in a 5G Mobile Packet Core Environment Smart Traffic Protection in TSN Networks Based on Application Needs •Selective Traffic Protection Activation in TSN Networks (using FRER) Based on Application Type 18 Experiments DetNet PREOF 19 Packet Replication, Elimination and Ordering Functions (PREOF –DetNet) 100G 10G DetNet domain IP domain CSP-7550 P4 - 2 CSP-7550 P4 - 3 CSP-7550 P4 –1 +FPGA DCSG_3 TG Spirent Impairment device IXIA NE2 Push DetNet & Replication Forwarding Elimination DetNet domain LILIANA - DPDK Reorder & Pop DetNet 1 30X 12 3 1 23 1 23 23 123312 20 Experiments Basic characterization of FlexE performance 21 10 GE round trip delay –IP vs FlexE Tester Round trip delay is 4x the value of one way delay in a single device ~ 15 µs ≈ 3 km 22 10 GE jitter - IP vs FlexE Cumulative jitter (4 hops) Tester 23 Experiments Characterization of critical applications 24 Characterization of critical applications •Impact of network conditions in Smart factory environments •To put into perspective how the network conditions such as latency, jitter, packet ordering, or packet loss would impact in the Smart Factory scenario •Characterization follows ITU-T G.1051 specification 25 Work in Progress Path Computation based on Precision Metrics