Access Official Technical PDF

Share Summary:LinkedInEmail

Industrial applications that depend on predictable data delivery can expose a limitation of conventional IP networking: best-effort forwarding does not guarantee a particular flow’s bandwidth or latency. In its report on next-generation transport technologies, ETSI’s Next Generation Protocols Industry Specification Group outlines a framework for adding IP quality of service (QoS) while retaining core IP architecture and supporting existing transport protocols.

The proposal is relevant to automation environments where applications may have different communications needs. Rather than expecting every TCP or UDP flow to adapt to the same best-effort network, the framework considers how the network itself could provide differentiated forwarding services, coordinated with transport-layer behavior.

Why best-effort IP can constrain transport performance

In a traditional IP network, routers generally forward traffic without guaranteeing an individual flow’s throughput, packet loss, latency, or jitter. TCP and applications compensate through mechanisms such as flow control and congestion control. These mechanisms help manage shared resources, but they cannot remove the constraints imposed by the underlying network.

TCP congestion-control algorithms infer network conditions and adjust the amount of data sent. Their behavior can depend on factors including round-trip time, packet loss, competing traffic, and available capacity. A flow’s actual throughput is therefore not guaranteed. Network fairness policies can also limit a flow to its share of a congested link, even when an application would benefit from more bandwidth.

Latency presents a related challenge. Approaches that increase link utilization can contribute to queue buildup, while efforts to reduce queuing may affect throughput. Active queue management and other congestion-control techniques can help address aspects of this trade-off, but the report notes that they do not necessarily provide deterministic latency for a specific flow.

UDP does not provide TCP-style flow or congestion control by itself. As a result, UDP applications also depend on network behavior and may need their own congestion-management mechanisms. The underlying issue is broader than a particular transport protocol: best-effort forwarding does not offer per-flow QoS guarantees.

IP quality of service needs network participation

ETSI’s framework proposes evolving best-effort IP toward a network that can support multiple services, including bandwidth guarantees or bounded-latency services. The report uses the term deterministic IP for an IP service intended to provide QoS for bandwidth and minimum latency. It presents this as a framework for development, not as a claim that ordinary IP networks already provide those guarantees.

A central design principle is to involve network devices more directly in transport performance. Instead of relying exclusively on endpoints to infer congestion, the framework considers how network nodes can establish and apply QoS forwarding behavior for selected traffic. Existing IP architecture, routing, and packet forwarding are intended to remain in place, while additional capabilities would enable differentiated service.

The report distinguishes this approach from established QoS models. DiffServ manages traffic in aggregated classes and is typically configured within service-provider networks. IntServ can support finer-grained reservations, but maintaining per-session state can be difficult as network scale grows. MPLS traffic engineering can provide differentiated services, but relies on additional mechanisms and is not generally configured from end-user applications. The proposed framework examines a different way to coordinate flow-level requirements with IP forwarding.

How in-band signaling fits the framework

In-band signaling carries control information within the same channel as user data. In the ETSI proposal, signaling could communicate information about a flow’s requested service and the forwarding state needed to support it. The framework discusses a network control plane as well as a data plane, with messages and parameters for setting up and reporting service state.

For IPv6, the report examines using extension-header-based signaling and describes control scenarios for TCP. Its design topics include:

  • Service setup: communicating the requested QoS and tracking setup status.
  • Flow identification: identifying the traffic to which a forwarding treatment applies.
  • Forwarding state: conveying and reporting the state used by network devices to forward traffic.
  • Operations and management: supporting monitoring and handling of service or forwarding conditions.
  • Security: considering authentication and protection of control mechanisms.

The framework also discusses IPv4, host behavior, and other transport protocols. Its detailed design focus is TCP, so the report should not be read as a complete implementation specification for every protocol or network environment.

Congestion control and operational trade-offs

Network-provided QoS would not eliminate the need for congestion control. Hosts still need to manage transmission, and the network must handle resource allocation and traffic management. The report therefore treats host congestion control and network forwarding as related parts of the overall solution.

Differentiating flows also changes the assumptions behind conventional fairness. Best-effort networks commonly aim to share resources fairly among competing traffic. Providing selected flows with distinct bandwidth or latency treatment means that service-provider or network-operator policies must determine how resources are partitioned between differentiated and traditional services. The report emphasizes that ordinary fairness-based service remains useful for many applications and can coexist with new QoS services.

For plant and network teams, this distinction matters in design reviews. A QoS request needs to be connected to a defined service, identifiable traffic, network support along the path, and operational controls. A host-side transport change alone cannot create an end-to-end guarantee if the network does not recognize and honor the requested treatment.

Practical implications for automation networks

ETSI’s framework offers a way to think about transport requirements that go beyond average throughput. Engineers evaluating demanding communications should consider whether an application needs bandwidth assurance, bounded latency, or simply best-effort connectivity—and whether the network can support that requirement across the relevant path.

The report also highlights the implementation work involved: network hardware and software must support the required signaling and forwarding behavior; hosts and applications must interact appropriately with the service; and operators need policies for resource allocation, monitoring, and security. These are architectural considerations, not capabilities that can be assumed from the presence of TCP or IP alone.

Conclusion

The ETSI proposal reframes transport performance as a shared responsibility between endpoints and the IP network. By combining network control, in-band signaling, differentiated forwarding, and host congestion management, a QoS-enabled IP network could address some limits of best-effort transport for applications with demanding bandwidth or latency needs. For industrial decision-makers, the practical takeaway is to define service requirements clearly and assess the complete network path, rather than relying on transport-layer tuning alone to deliver predictable performance.

Access Official Technical PDF

About The Author