![Architecture Model
ES1
ES2
BR1 ES3
r1
r2
(a) Example architecture
0
for TT traffic in ES1 is open, therefore the TT frame of tTT1
is initiated to be transmitted from [ES1,SW1].
Researchers have proposed methods to synthesize the G-
CLs [14], [24], and have outlined the constraints that have
to be satisfied for schedule feasibility. For example, when
associated gate for TT traffic is open, the remaining gates
for other traffic are closed, and vice versa. In Fig. 3, the red
and blue queues are respectively dedicated for Class A and
Fig. 2. A TAS for an output port in an ES/SW
B o
TT
Clas
from
is c
The
tran
A. T
H
sion
TT
3
(b) TSN switch
4 / 9](https://image.slidesharecdn.com/slides-180707150430/85/Scheduling-in-Time-Sensitive-Networks-TSN-for-Mixed-Criticality-Industrial-Applications-4-320.jpg)
Scheduling in Time-Sensitive Networks (TSN) for Mixed-Criticality Industrial Applications
- 1. Scheduling in Time-Sensitive Networks (TSN) for Mixed-Criticality Industrial Applications Voica Gavrilut¸ Paul Pop Technical University of Denmark (DTU) DTU Compute Kongens Lyngby, Denmark (voga—paupo)@dtu.dk
- 2. Motivation The Automation Pyramid: Soon to be Ancient History? 22 Rigid infrastructure with separation between levels of functionality Levels connected by dedicated, specialist networks Data exchange only via gateways or proprietary systems Difficulties to transparantly access data at the cyber pyramid (machine) level (a) Automation Pyramid Creating a Flexible IoT Infrastructure with Deterministic Ethernet 24 OPC UA over TSN: one unified architecture for all communication infrastructure elements Logical machine/control boundaries are dissolved Direct access to machine data from ERP/MES Real-time and non-real-time domains integrated OPC UA provides leading technology of proven security concepts Learning from the Internet Experience: Digital Platforms and Open Innovation win. Source: TTTech (b) Distributed CPSs Figure Specialized protocols as ProfiNet or EtherCAT replaced by TSN Single infrastructure supporting mixed-criticality applications 2 / 9
- 3. Time-Sensitive Networking (TSN) TSN: a set of substandards extending IEEE 802.1Q Shapers and schedulers: IEEE 802.1Q-2005: prioritized best-effort IEEE 802.1BA: Audio-Video-Bridging (AVB) IEEE 802.1Qbv: enhancements for Time-Triggered (TT) traffic our focus: mixed AVB and TT traffic IEEE 802.1Qat, IEEE 802.1Qcc: stream reservation IEEE 802.1CB: stream reservation for redundant streams 3 / 9
- 4. Architecture Model ES1 ES2 BR1 ES3 r1 r2 (a) Example architecture 0 for TT traffic in ES1 is open, therefore the TT frame of tTT1 is initiated to be transmitted from [ES1,SW1]. Researchers have proposed methods to synthesize the G- CLs [14], [24], and have outlined the constraints that have to be satisfied for schedule feasibility. For example, when associated gate for TT traffic is open, the remaining gates for other traffic are closed, and vice versa. In Fig. 3, the red and blue queues are respectively dedicated for Class A and Fig. 2. A TAS for an output port in an ES/SW B o TT Clas from is c The tran A. T H sion TT 3 (b) TSN switch 4 / 9
- 5. Application Model flow type vs vt r T D P f1 · · · f4 TT ES1 ES3 r1 150 µs 150 µs 750 B f5 · · · f8 AVB ES2 ES3 r2 150 µs 150 µs 1500 B 1.1 3.12.1 4.1 8.1 6.17.1 5.1 2.1 4.18.1 6.1 1.1 3.17.1 5.1 50 100 150 50 100 150 50 100 150 ES1, SW1 ES2, SW1 SW1, ES3 6, 8 7 58 6 151.76 us 5 Figure: GCL for an outgoing port: WCD(f5) = 151.76µs 5 / 9
- 6. Problem Formulation Given: The network G and The set of flows F = FTT ∪ FAVB with their routes and timing properties Determine: The number of TT queues, Mapping of TT flows to egress port queues and GCLs S of TT queues Such that: All TT flows are schedulable Objective function: Cost = fi ∈FAVB max(0, WCD(fi ) − fi .D) 6 / 9
- 7. Optimization Strategy NP-hard problem TT+AVB: metaheuristic strategy Metaheuristic: Greedy Randomized Adaptive Search Procedure (GRASP) Has 2 phases: 1. Construction phase; generate a list of candidates using variations of List Scheduling heuristic 2. Search phase; improvement of each candidate by applying destroy/repair operators. 7 / 9
- 8. Preliminary Experimental Results TT latency TT Queues TT+AVB Test case Exec. Sched. Exec. Sched. Exec. Sched. Motiv. 1.1 No 0.8 No 32 Yes TC1 2.2 No 2.6 No 403.9 Yes TC2 8.7 No 9.6 No 612.8 Yes 8 / 9
- 9. Closing Remarks Summary Addressed the scheduling of mixed-criticality applications on TSN-based networks Problem: decide the GCLs of TT queues Solution: a GRASP-based metaheuristic Message Is not obvious what schedule tables are more convenient also for AVB traffic We need scheduling tools to (1) Synthesize GCLs in TSN and (2) Modify the GCLs such that also AVB traffic requirements are met Future work Determine the GCLs for multicast flows Modify also the routes Modify also the idle/sending slopes of AVB classes 9 / 9