The TSSP GaZ’s team addresses the design, analysis, and verification of systems with temporal constraints, in which correctness depends not only on the logical result but also on the instant at which it occurs — the property that defines Real-Time Systems (RTS). Although RTS constitute its classical core, the timing measurement and characterization, formal modeling, and emulation techniques developed by the team go beyond the traditional problem of schedulability and are applicable to time-sensitive systems in a broader sense. An example is the emulation of a TSN system, bridging time-aware and non-time-aware networks or fine-tuning frameworks such as ROS2. Application domains include industrial automation, in-vehicle systems, and aerospace platforms, where flows of heterogeneous criticality must coexist over shared communication infrastructures with deterministic latency and jitter guarantees.
Research lines
- Scheduling and routing in deterministic networks (TSN). Time-Sensitive Networking (IEEE 802.1) extends Ethernet with determinism mechanisms, including the Time-Aware Shaper (IEEE 802.1Qbv), whose configuration requires solving the joint routing and schedule synthesis problem (Routing & Scheduling), which is NP-complete in nature. Integer linear programming formulations (ILP/MILP) are developed for its optimal resolution, both global and incremental (integrating new flows without disrupting the ongoing schedule), with multipath support and fault-tolerance mechanisms based on replication over disjoint paths (FRER, IEEE 802.1CB).
- Performance analysis on real platforms. The timing guarantees of a schedule depend critically on the variability introduced by the processor, the memory hierarchy, the network controller, and the operating system. The research line empirically characterizes these sources of indeterminism on x86 and ARM platforms under Linux PREEMPT_RT, including the study of low-latency waiting primitives and their trade-offs between latency, precision, and consumption in real-time systems, as well as the performance analysis of eBPF programs in the kernel’s packet processing paths.
- Formal modeling and analysis. As a complement to the experimental approach, mathematical formalisms, such as Petri nets (including their timed and continuous extensions) and network calculus, are used to model TSN networks, derive delay and backlog bounds, and analyze the structural and control properties of traffic-shaping mechanisms.
The group operates a testbed comprising five end stations and five bridges, currently under expansion.
Team
Senior researchers
- Department of Computer Science and Systems Engineering / I3A, University of Zaragoza:
- Prof. José Luis Briz, PhD
- Prof. Juan Segarra Flor, PhD
- Center for Research and Advanced Studies (CINVESTAV) Guadalajara, Mexico:
- Prof. Antonio Ramírez-Treviño, PhD
Industry collaborators
- Héctor Blanco-Alcaine, MSc (Intel Corporation | Intel Deutschland GmbH)
PhD students
- Alitzel Torres-Macías (CINVESTAV – University of Zaragoza)
- Álex Gracia Rodríguez (University of Zaragoza)
The group also supervises a selection of final-degree and master’s students, who receive specific training in the technologies described above and typically express interest in continuing their work in this area.
Recent publications
- A. G. Torres-Macías, Á. Gracia, J. Segarra, J. L. Briz, A. Ramírez-Treviño and H. Blanco-Alcaine, «Optimal Fast IEEE802.1Qbv Incremental Scheduling,» in Ad Hoc Networks, 2026, Art. no. 104280, ISSN 1570-8705, doi: 10.1016/j.adhoc.2026.104280.
- A. Galilea Torres-Macías, J. Segarra Flor, J. Luis Briz Velasco, A. Ramírez-Treviño and H. Blanco-Alcaine, «Fast IEEE 802.1Qbv Gate Scheduling Through Integer Linear Programming,» in IEEE Access, vol. 12, pp. 111239–111250, 2024, doi: 10.1109/ACCESS.2024.3440828.
- Á. Gracia, A. G. Torres-Macías, J. Segarra, J. L. Briz, A. Ramírez-Treviño, H. Blanco-Alcaine. «Embedded reconfiguration of TSN: Dual reconfiguration with dropping and reclaiming.» 7th Euromicro Conference on Real-Time Systems (ECRTS) – Industrial Challenge. July 8–11, 2025, Brussels, Belgium.
- Álex Gracia, José Luis Briz, Héctor Blanco-Alcaine, Juan Segarra, Alitzel G. Torres-Macías, and Antonio Ramírez-Treviño. 2025. «Characterization of latency and jitter in TSN emulation.» arXiv preprint arXiv:2506.02133.
- A. Gracia, J. L. Briz Velasco, H. Alcaine, J. Segarra, A. Torres, A. Ramírez-Treviño. «Cracking down overheads in TSN emulation over Mininet.» Time-Sensitive Networking and Applications (TSN&A). Stuttgart, 1–2 Oct. 2024.
- A. G. Torres-Macías, A. Ramírez-Treviño, J. L. Briz, J. Segarra, H. Blanco-Alcaine. «Modeling Time-Sensitive Networking Using Timed Continuous Petri Nets.» IFAC-PapersOnLine, Volume 58, Issue 1, 2024, Pages 300–305, ISSN 2405-8963. https://doi.org/10.1016/j.ifacol.2024.07.051
Collaboration opportunities
The group is open to collaboration in two broad modalities:
- Direct application of the group’s expertise to specific technical problems — particularly profiling, fast-packet processing, and any aspect amenable to performance acceleration through kernel and hardware configuration or techniques such as eBPF/XDP and AF_PACKET.
- Formal industry–research partnerships, which may involve test or demonstration infrastructures for key enabling technologies (TSN and related), or technology transfer activities ranging from proof-of-concept to prototype and higher TRL levels.
A quick dive into Time-Sensitive Networks (TSN)
Main components
A TSN network consists of two principal node types: end stations and bridges (switches). Both are synchronized via the IEEE 802.1AS profile of the IEEE 1588 PTP specification, achieving synchronization on the order of nanoseconds. End stations — PLCs, servers running vPLCs, and similar devices — originate and terminate traffic, transmitting information (e.g., from sensors to actuators) under guaranteed timing constraints in terms of both deadline and jitter.
Domain requirements and traffic classes
Latency and jitter requirements have tightened progressively across application domains to accommodate increasingly demanding use cases:
| Sector | Application | Latency | Jitter |
|---|---|---|---|
| Health | Tele-Surgery, Haptic Feedback | 3–10 ms | < 2 ms |
| Manufacturing | Industrial Automation, Control Systems | 0.2 µs–0.5 ms (1 Gbit/s); 25 µs–2 ms (100 Mbit/s) | Meet latency req. |
| Energy | Power Grid Systems | ≈ 8 ms | Few µs |
| Banking | High-Frequency Trading | < 1 ms | Few µs |
| Aerospace | AFDX Variants | 1–128 ms | Few µs |
| Automotive and transport | Advanced Driver Assistance Systems (ADAS) | 100–250 µs | Few µs |
| Automotive and transport | Power Train, Chassis Control | < 10 µs | Few µs |
| Automotive and transport | Traffic Efficiency & Safety | < 5 ms | Few µs |
| Infotainment | Augmented Reality | 7–20 ms | Few µs |
| Professional Audio/Video | — | 2–50 ms | < 100 µs |
To address this diversity, TSN defines six domain-specific profiles:
- IEEE P802.1DP — Aerospace onboard Ethernet
- IEEE/IEC 60802 — Industrial automation
- IEEE P802.1DG — Automotive in-vehicle communications
- IEEE P802.1CM — Fronthaul
- IEEE P802.1DF — Service provider networks
- IEEE P802.1BA — Audio/video bridging (AVB)
Each profile defines traffic classes relevant to its sector. The industrial automation profile (IEC/IEEE 60802) distinguishes six:
- Latencies below 2 ms, no congestion loss, strict synchronization, fixed frame sizes (30–100 bytes).
- Periodic, time-sensitive traffic with latencies between 2 and 20 ms and defined bandwidth requirements.
- Aperiodic traffic — alarms, operator commands, critical control events — with frame sizes of 100–1500 bytes.
- Highly critical periodic traffic supporting network management functions such as gPTP and SRP.
- Sporadic, medium-criticality traffic without strict temporal constraints (diagnostics, logs).
- Non-critical traffic with no timing constraints.
Bridges and traffic management
Bridges enforce prioritized transmission through traffic shapers and dynamic scheduling aligned with the traffic classes of the applicable profile. IEEE 802.1Q supports the dynamic admission of new flows via a hitless set-and-hold mechanism — modifications are staged in an administrative state while the network continues operating under the current operational state, then atomically committed at a designated cycle boundary without disrupting ongoing transmissions.
In practice, configuration is typically centralized. A Centralized User Configuration node (CUC) defines the network topology and communication flow requirements, and passes this specification to a Centralized Network Configuration node (CNC), which is responsible for planning, verification, and deployment. IEEE 802.1Q further encompasses specifications for network management, fault tolerance, and quality of service.
Links and contact
Prof. José Luis Briz
DIIS/I3A – EINA, Univ. Zaragoza
briz@unizar.es