PaperlayerView exact revision on arXiv

Source-linked AI summary

Ultra-Low Latency (ULL) Networks: The IEEE TSN and IETF DetNet Standards and Related 5G ULL Research

Ahmed Nasrallah, Akhilesh Thyagaturu, Ziyad Alharbi, Cuixiang Wang, Xing Shao, Martin Reisslein, Hesham ElBakoury

arXiv:1803.07673v3cs.NI

TL;DR

Many ULL applications outstrip the latency capabilities of traditional packet networks. This paper surveys TSN, DetNet, and related 5G ULL standards and research across major flow mechanisms, identifies their limitations, and outlines future directions. The survey presents TSN and DetNet as complementary link-layer and network-layer approaches, while emphasizing unresolved interoperability, management, synchronization, and reliability challenges.

  • Problem

    Traditional Ethernet and packet networks lack deterministic end-to-end QoS, while ULL applications require bounded latency, low jitter, low loss, and reliable flows.

  • Method

    The paper conducts a comprehensive survey of TSN, DetNet, and 5G ULL standards and research, organizing mechanisms by flow properties and 5G network segment.

  • Results

    The survey covers TSN link-layer mechanisms, DetNet network-layer specifications, and 5G fronthaul, backhaul, and management studies, including their identified pitfalls and limitations.

  • Takeaways & Limitations

    TSN and DetNet provide complementary deterministic support across link and network layers, while 5G ULL support depends on integrating these mechanisms across the wireless access chain.

  • Takeaways & Limitations

    TSN research has largely focused on isolated in-vehicle networks, with limited large-scale heterogeneous-network simulation and insufficient standards for connecting multiple TSN networks.

Abstract

from arXiv · show

Many network applications, e.g., industrial control, demand Ultra-Low Latency (ULL). However, traditional packet networks can only reduce the end-to-end latencies to the order of tens of milliseconds. The IEEE 802.1 Time Sensitive Networking (TSN) standard and related research studies have sought to provide link layer support for ULL networking, while the emerging IETF Deterministic Networking (DetNet) standards seek to provide the complementary network layer ULL support. This article provides an up-to-date comprehensive survey of the IEEE TSN and IETF DetNet standards and the related research studies. The survey of these standards and research studies is organized according to the main categories of flow concept, flow synchronization, flow management, flow control, and flow integrity. ULL networking mechanisms play a critical role in the emerging fifth generation (5G) network access chain from wireless devices via access, backhaul, and core networks. We survey the studies that specifically target the support of ULL in 5G networks, with the main categories of fronthaul, backhaul, and network management. Throughout, we identify the pitfalls and limitations of the existing standards and research studies. This survey can thus serve as a basis for the development of standards enhancements and future ULL research studies that address the identified pitfalls and limitations.

I. INTRODUCTION

ULL applications require latency from a few microseconds to milliseconds, while traditional packet networks typically achieve only tens of milliseconds. This survey organizes prior standards and research on TSN, DetNet, and 5G ULL, while identifying gaps for future work.

  • Motivation: Traditional packet networks reduce end-to-end latency only to tens of milliseconds, whereas ULL applications require far shorter bounds.Industrial applications may require a few microseconds to a few milliseconds, tactile Internet applications around 1 millisecond, and cellular fronthaul around 100 microseconds.
  • Survey scope: The survey provides an up-to-date review of IEEE TSN standards, related TSN research, IETF DetNet standards, related DetNet research, and 5G ULL studies.Its organization covers flow concepts, synchronization, management, control, and integrity, with 5G studies grouped into fronthaul, backhaul, and network management.
  • Survey scope: The paper identifies limitations and research gaps in existing TSN, DetNet, and 5G ULL standards and studies, then outlines future research directions.It also differentiates this work from prior surveys covering general latency reduction, specific protocols, the Tactile Internet, or AVB.
  • Motivation: ULL commonly means latency of a few milliseconds or less than one millisecond, often with deterministic bounds for every frame in a flow.Some multimedia applications instead require probabilistic latency guarantees, where rare violations have negligible perceived impact.
  • Motivation: Ethernet offers simple, high-speed connectivity but fundamentally lacks deterministic end-to-end QoS, including guaranteed bandwidth and policy enforcement.These shortcomings motivated IEEE and IETF efforts to introduce deterministic packet-flow concepts.

III. IEEE TSN STANDARDIZATION

IEEE TSN extends Ethernet with mechanisms for synchronized, managed, controlled, and reliable low-latency flows. Its standardization is organized around TSN flow properties and has evolved from AVB through continuing revisions.

  • TSN Data-Link Flow: A TSN flow is an end-to-end unicast or multicast connection between talkers and listeners across a time-sensitive network.The flow is characterized by QoS properties such as bandwidth and latency, and identified using the PCP and VID fields in the Ethernet VLAN tag.
  • TSN organization: The survey classifies TSN standardization by flow properties and supplements that taxonomy with a timeline of standards milestones and AVB-to-TSN evolution.The classification covers flow characterization, synchronization, management, control, and integrity mechanisms.
  • Network Timing: IEEE 802.1AS provides network-wide time synchronization by exchanging timing information through a hierarchy centered on a selected grandmaster clock.Accuracy depends mainly on residence-time and link-delay measurements; logical syntonization corrects timing using the grandmaster’s frequency ratio.

C. Flow Management

TSN flow management supports distributed and centralized provisioning, reservation, monitoring, and configuration of network resources. These mechanisms create protected channels over shared networks while addressing scalability limits of decentralized registration.

  • Flow management dynamically discovers, configures, monitors, and reports bridge and end-station capabilities.
  • IEEE 802.1Qat Stream Reservation Protocol: IEEE 802.1Qat admits or rejects flows according to resource requirements and available network resources, then reserves resources and advertises streams across Ethernet links.
  • IEEE 802.1Qat Stream Reservation Protocol: MRP registers stream attributes across bridges, while MSRP distributes reservations and QoS parameters from talkers to network devices.
  • IEEE 802.1Qat Stream Reservation Protocol: Decentralized SRP registration can overwhelm the network when registrations or de-registrations change, causing intolerable delays for critical traffic classes.
  • IEEE 802.1Qcc Enhancements: IEEE 802.1Qcc adds a UNI and CNC for centralized resource reservation, scheduling, configuration, and management, while supporting distributed and hybrid configurations.
  • Scalability limitations: MRP’s approximately 1500-byte stream-state database limits scalability as traffic streams and network size increase.

D. Flow Control

TSN flow-control standards shape transmission according to traffic class, using credit-based and time-aware mechanisms to bound delay and protect scheduled traffic. CBS provides bounded latency but can incur high per-hop delay, motivating TAS, preemption, and asynchronous shaping alternatives.

  • IEEE 802.1Qav Forwarding and Queuing: IEEE 802.1Qav provides bounded-latency, bounded-jitter, lossless forwarding and queuing for time-sensitive audio/video traffic.
  • IEEE 802.1Qav Forwarding and Queuing: Credit-Based Shaping spaces frames to reduce bursting, using non-negative credit eligibility with idleSlope increases and sendSlope decreases during transmission.
  • IEEE 802.1Qav Forwarding and Queuing: Under 2 ms for class A and under 50 ms for class B up to seven network hops are the IEEE 802.1Qav worst-case latency guarantees.
  • IEEE 802.1Qav limitations: CBS can add up to 250 µs per hop and struggles to maintain delay guarantees at high link utilization, limiting its suitability for some industrial control applications.
  • IEEE 802.1Qbv Time-Aware Shaper: TAS schedules high-priority traffic in synchronized time windows through gate control lists, while frame preemption can reduce the guard band protecting those windows.
  • IEEE 802.1Qbv Time-Aware Shaper: TAS requires synchronized bridges and incurs sampling delay when data arrives outside the next scheduled window, motivating coordination between end-node and network scheduling.
  • IEEE 802.1Qcr Asynchronous Traffic Shaping: Asynchronous Traffic Shaping’s UBS design targets predictable worst-case delays at high utilization, lower implementation complexity, and independence from global time synchronization.

6) Summary and Lessons Learned:

TSN combines flow prioritization, shaping, reservation, path control, filtering, redundancy, and security-related mechanisms to support deterministic low-latency communication. The standards provide broad QoS and reliability capabilities, but dynamic configuration, jitter control, security integration, and industry validation remain open challenges.

  • Flow control and integrity: TSN flow control prioritizes and queues frames by traffic class, while flow management and integrity mechanisms help ensure adequate resources and reliable delivery.Existing controls privilege TSN flows over non-TSN traffic; flow control collaborates with management and integrity functions.
  • Flow control and integrity: IEEE 802.1Qbv and Qcr represent time-triggered and event-triggered shaping, motivating research on combining flexible runtime adaptation with deterministic latency guarantees.Event-triggered shaping may offer flexible schedules, whereas replacing TAS could guarantee an upper latency bound without generally guaranteeing deterministic latency.
  • Flow integrity: 802.1CB FRER replicates critical frames over disjoint paths and eliminates duplicates at the destination to limit the effects of congestion and link faults.Replication can use stream identification and sequence numbers to determine which copies to discard and which to forward.
  • Flow management and integrity: IEEE 802.1Qca supports explicit multipath forwarding through IS-IS and SDN-assisted control, while 802.1Qci filters and polices individual streams to prevent overload and attacks.Qca configures protected paths; Qci identifies streams with StreamID and processes them through filtering, aggregation, and gating.
  • TSN standardization: TSN standardization provides Layer 2 deterministic networking with QoS, low latency, jitter and loss support, redundancy, reservations, bandwidth limitation, reconfiguration, centralized management, and strict timing.The standards also include mechanisms for dynamic reconfiguration and resource reservation, but efficient dynamic configuration remains an open challenge.
  • Open limitations: TSN does not explicitly control jitter, security and privacy integration remains incomplete, and significant testing and benchmarking are still needed for industry and consumer assurances.CQF can reduce or bound delay-related jitter, but dedicated jitter throttling and integrated TSN security remain future issues.

IV. TSN RESEARCH STUDIES

TSN research studies address synchronization, resource reservation, reconfiguration, scheduling, routing, shaping, preemption, and fault tolerance. The studies show promising hardware and synchronization results while exposing overhead, accuracy, scalability, and runtime-management challenges.

  • Research organization: TSN research is organized around synchronization, resource reservation and reconfiguration, bandwidth allocation, routing, scheduling, shaping, preemption, and fault tolerance.The survey presents these categories as the main areas of TSN-related ULL research.
  • Synchronization: Most existing time-synchronization implementations achieve clock precision on the order of sub-microseconds, but network-wide synchronization depends on diverse hardware and network attributes.The survey identifies clock precision, frequency synchronization, and timing accuracy as distinct research topics.
  • Synchronization: A synchronized FPGA/x86-64 architecture achieved cut-through latencies of 2 to 2.5 µs for twelve Gigabit Ethernet ports at full-line-rate packet processing.The reported latency included pipeline, arbitration, aggregation, backpressure, clock-domain, width-adaptation, and head-of-line-blocking delays.
  • Synchronization: Synchronization traffic imposes bandwidth and control-plane overhead, while blindly following inaccurate grandmaster timing can produce out-of-sync clocks.Research considers timing-error estimation and accuracy checks using received timing information or tracked node and port errors.
  • Resource reservation: Automatic reservation de-registration reduces stale resource occupancy through synchronized idle-time monitoring, but short inactive periods can trigger costly reactivation for bursty traffic.The approach was proposed because distributed reservation release can create substantial control-message exchanges in larger networks.
  • Resource reservation: Runtime reconfiguration of IEEE 802.1Qbv gate-control schedules remains difficult: a greedy earliest-deadline-first approach required several seconds to up to a minute.The survey identifies dynamic runtime management and reconfiguration as significant future research challenges.

2) Bandwidth Allocation:

TSN bandwidth allocation and scheduling reserve and shape resources to bound latency, but optimization, adaptability, and mixed-traffic support remain challenging. Research improves schedules and routing while exposing tradeoffs among efficiency, complexity, and traffic classes.

  • Bandwidth allocation: Bandwidth allocation reserves transmission resources for end-to-end delay requirements across multiple TSN traffic classes.These classes include scheduled traffic with different priorities and best-effort non-scheduled traffic.
  • Routing: Optimization-based routing minimizes TSN path delays while considering multipath jitter and loop probability.
  • Flow management: Dynamic flow management can improve efficiency through statistical multiplexing, but adapting to topology and node changes requires further control-plane research.Traffic shaping adds queuing delay, whereas policing can drop excess frames and reduce TCP throughput.
  • Scheduling: IEEE 802.1Qbv does not specify schedule-computation algorithms, and its guard bands can waste bandwidth and increase latency.Future work targets fewer guard-band occurrences while preserving traffic isolation.
  • Scheduling: TSN scheduling research computes near-optimal schedules for more than 1500 flows while reducing guard-band occurrences by 24%.The optimization also reduces overall end-to-end latency, but can increase node complexity and operational cost.
  • Traffic scheduling: TSN benefits primarily scheduled flows, while mixed best-effort traffic lacks guaranteed effective behavior and requires integrated worst-case analysis.Studies also examine bounded latency and jitter under varying loads, scheduling methods, and network configurations.

3) Preemption:

Preemption prioritizes time-critical TSN traffic and can substantially reduce its latency, but it shifts delay and complexity to lower-priority traffic. Reliable deployment also depends on synchronization, redundancy, interoperability, and implementation choices.

  • Preemption effects: Preemption prioritizes CDT frames over regular Ethernet traffic, producing different delay effects across traffic priorities.Lower-priority CDT frames can also be preempted by higher-priority CDT frames.
  • Preemption effects: 60% lower worst-case CDT latency was reported with preemption than with baseline 802.1Q without preemption.Worst-case non-CDT latency increased up to 6% because of preemption overhead.
  • Preemption analysis: Express traffic average frame delays were one to over three orders of magnitude shorter than preemptable traffic in queueing-model and simulation evaluations.The study also provided VHDL transmit- and receive-unit designs for FPGA implementation.
  • Fault tolerance: TSN fault tolerance uses frame replication, elimination, path control, and reservations, but recovery must preserve end-to-end behavior and network topology.Research on resource reservation for fault tolerance under ULL requirements remains scant.
  • Fault tolerance: Redundancy choices trade flexibility against protocol overhead and bandwidth, with decoupled reservation and redundancy better suited to some peculiar industrial requirements.The appropriate approach depends on application requirements.
  • Interoperability: TSN’s closed-network focus makes interoperability with external networks essential for heterogeneous robotic and vehicular deployments.Interconnected TSN networks can use wireless protocols and shared timing information to establish a common timing platform.

5) Summary and Lessons Learnt:

The survey identifies interoperability, scalability, management complexity, and timing overhead as central TSN and DetNet challenges. It also frames DetNet as a layer-3 extension of TSN whose standardization remained incomplete in the surveyed period.

  • TSN research lessons: TSN research must address interoperability across heterogeneous LAN, WAN, core, cellular, and WLAN architectures rather than isolated closed networks.Existing work largely focused on in-vehicle networks, with limited large-scale heterogeneous simulation and external-network evaluation.
  • TSN research lessons: Future TSN designs should support applications ranging from time-sensitive to delay-tolerant traffic with flow-level scheduling and dynamic priority allocation.The stated requirements include bounded latency for lower-priority traffic and connectivity between multiple closed TSN architectures.
  • TSN research lessons: SDN-based centralized management, efficient timing sharing, local-clock skew correction, and computationally efficient hardware and software are identified as future TSN requirements.These requirements target global management, accurate timing, and reduced implementation cost.
  • TSN research lessons: Deterministic TDM provides a 100% latency bound but can increase average delay and reduce utilization for bursty traffic compared with statistical multiplexing.Statistical guarantees may trade rare bound violations for higher utilization and lower average delay.
  • DetNet standardization: DetNet extends TSN capabilities to layer-3 routed segments, but its early standardization relied on drafts and faced layering, identification, and architecture-complexity issues.DetNet introduced service and transport layers, multiple flow types, cross-layer awareness, and flow-attribute mapping requirements.

B. Flow Synchronization

DetNet synchronization and configuration support deterministic QoS across routed networks by combining timing principles, multiple control-plane models, capability dissemination, and path computation. The survey emphasizes that scalable admission control and standardized data-plane choices remained open issues.

  • Synchronization: DetNet seeks to bound minimum and maximum latency through sub-microsecond synchronization and time-of-execution fields embedded in application packets.These mechanisms target jitter reduction for mission-critical latency traffic.
  • Configuration: DetNet supports fully distributed, fully centralized, and hybrid configuration models, with a YANG model for conveying configuration parameters.The models distribute or centralize UNI and control information across DetNet entities.
  • Configuration: DetNet configuration requires topology, path, resource, node-capability, and neighbor-link information for dynamic management and path installation.The information includes queueing algorithms, buffers, forwarding delay, available resources, link length, and bandwidth.
  • Path setup: Centralized DetNet path setup uses PCEs and network information to optimize paths globally and per flow, while distributed setup uses IGP-TE signaling protocols.The distributed approach includes MPLS-TE, RSVP-TE, OSPF-TE, and ISIS-TE.
  • Resource management: DetNet must develop effective admission control and resource management for saturated networks so higher-priority ULL traffic receives appropriate resources.The surveyed framework does not yet standardize how capability information is carried and implemented across large-scale networks.
  • Data plane: No official DetNet network-layer data-plane encapsulation had emerged, leaving multiple service- and transport-layer protocol candidates under consideration.The candidate overview identifies UDP over IP as a prominent deployment candidate.

3) Queuing, Shaping, Scheduling, and Preemption:

DetNet flow control combines TSN queuing, shaping, scheduling, preemption, admission control, and resource protection to provide bounded latency and low loss across routed segments. Its larger scale introduces interoperability, overhead, and cross-domain QoS challenges, while replication improves integrity at a bandwidth cost.

  • Flow control: DetNet leverages TSN credit-based, time-gated, cyclic, asynchronous, and preemption mechanisms to regulate flows and protect deterministic latency and loss bounds.The mechanisms include 802.1Q, 802.1Qbv, 802.1Qch, 802.1Qcr, 802.1Qbu, and 802.3br.
  • Flow control: Synchronous DetNet flows use closely synchronized nodes and periodic schedules, whereas asynchronous flows use worst-case interference calculations based on packet and transmission properties.Asynchronous regulation considers maximum packet size, observational interval, and maximum transmissions during that interval.
  • Flow control: DetNet limits each flow’s transmission opportunities and reserves bandwidth to achieve deterministic operation, while unused reserved bandwidth may serve non-DetNet traffic.The design regulates consumed bandwidth rather than permitting unrestricted transmission opportunities.
  • Traffic engineering: DetNet’s SDN-oriented traffic engineering architecture enables dynamic flow control, measurement, management, fast recovery, and deterministic QoS bounds across multiple layer-2 segments.The architecture defines application, control, and network planes and integrates with lower-layer transport functionality.
  • Limitations: DetNet flow control is more complex and larger in scale than TSN control, creating challenges in interoperability, control-data overhead, and QoS guarantees across diverse layer-2 segments.Multi-vendor reservation consistency and contractual QoS service-level issues are also identified.
  • Flow integrity: PREF adds sequence information, replicates flows over disjoint paths, and eliminates duplicates to reduce or nullify packet loss.A single-path variant can provide edge-node error correction without disjoint paths but requires extra bandwidth and suits low-rate traffic.

6) Security and Privacy Considerations:

DetNet’s convergence of IT and OT makes flow integrity, security, and privacy important, while packet replication improves reliability but introduces bandwidth and management costs. Existing DetNet research remains limited, leaving several architectural and protocol gaps.

  • Security and Privacy: DetNet integrates IT and OT networks, increasing the importance of security, privacy, authentication, and access control for converged cyber-physical systems.The survey identifies protection of signaling, authentication and authorization, and secure information dissemination as key concerns.
  • DetNet Mechanisms: DetNet uses centralized path computation and established redundancy mechanisms, including PCE, HSR, and PRP, to support deterministic flow navigation and reliability.The architecture combines IEEE and IETF standards with centralized SDN-based management.
  • Research Gaps: DetNet research has mainly addressed flow scheduling and replication, while control-plane management, virtualization, external-network interoperation, and broader architectural improvements remain open areas.DetNet also depends on TSN for deterministic Layer 2 synchronization and control features that it does not itself provide.
  • Flow Integrity: Packet replication can eliminate packet loss and reduce end-to-end latency by up to 40% with a replication factor of one.Higher replication factors substantially increase energy consumption, reaching approximately 2.914 times the baseline at factor four.
  • Limitations: Replication competes with ordinary traffic for bandwidth, potentially increasing congestion and delays, while additional replicated flows complicate failure-driven resource reallocation.The survey identifies low-latency coding as a possible alternative when mission-critical traffic occupies a large share of applications.

2) IEEE 802.1CM: Time-Sensitive Networking for Fronthaul:

IEEE 802.1CM specifies TSN mechanisms for carrying CPRI and eCPRI fronthaul over bridged Ethernet, using scheduling, preemption, synchronization, and traffic profiles to control latency. Its applicability is constrained by limited generalized-fronthaul support and unresolved security and reliability concerns.

  • Standard Scope: IEEE 802.1CM provides bridged Ethernet connectivity for CPRI and eCPRI fronthaul, supporting multiple radio functional splits.The standard requires at least 1 Gbps on each bridge port and defines mechanisms for end stations, bridges, and LANs.
  • ULL Mechanisms: IEEE 802.1CM uses scheduling, frame preemption, and synchronization to satisfy fronthaul latency requirements while enabling reuse of existing bridged networks.Centralized management can support automatic reconfiguration and reduce operational expenditures compared with manual configuration.
  • Latency Components: Bridge latency comprises store-and-forward, interference queueing, self-queueing, and periodic constant-bit-rate delays.Self-queueing depends on interfering gold flows arriving across multiple ingress ports and sharing an egress port.
  • Frame Preemption: Frame preemption lets a high-priority frame transmit after a preemptable fragment, reducing interference from a complete low-priority frame.Without preemption, queueing depends on the maximum low-priority frame plus preamble, SFD, and IPG; with preemption, the fragment includes CRC and IFG.
  • Limitations: IEEE 802.1CM primarily targets CPRI and eCPRI, so its lack of generalized-fronthaul support limits applicability to wider 5G scenarios such as crosshaul.Security and reliability were not considered in detail and remain areas for future standards and research.

3) Next Generation Fronthaul Interface (NGFI):

NGFI and related 5G core mechanisms pursue scalable, flexible transport and lower-latency backhaul by separating functional layers, control, and user-plane processing. Their deployment benefits depend on architecture, overhead, and operator cost.

  • NGFI Architecture: IEEE P1914.1 defines a two-level fronthaul architecture connecting RRH, DU, and CU through NGFI-I and NGFI-II interfaces.Alternative placements integrate CU and DU or integrate DU with RRH, producing different transport arrangements.
  • Design Goals: NGFI targets scalability, resource utilization, flexibility, cost effectiveness, and radio-technology-agnostic transport with SDN-controlled reconfiguration.It supports statistical multiplexing, heterogeneous networks, and multiple traffic classes with application-dependent latency requirements.
  • Limitations: The NGFI standard does not define fronthaul functional splits, limiting its architectural scope despite alignment with 3GPP for 5G-compatible splits.Its transport classes and interfaces support multiple split options, but functional split decisions remain outside IEEE 1914.1.
  • Backhaul Management: CUPS reduces cellular-network latency by placing user-plane nodes near the RAN and allowing data transport without control-plane interaction during path setup.User-plane paths can adapt dynamically to requirements and mobility.
  • Control-User Separation: CUPS separates control and user-plane functions, while retaining interfaces that communicate between the separated planes in the EPC.The architecture applies SDN-style control and data-plane separation to cellular backhaul core networks.

5) Next Generation (NG) Core:

The 5G NG Core decomposes EPC-style gateway functionality into service functions to improve deployment flexibility and support network management schemes. Across wireless access and backhaul, ULL research reduces transmission, processing, signaling, and path-management delays, but trade-offs remain.

  • NG Core Architecture: The NG Core separates network nodes into service functions for standalone 5G NR, while non-standalone 5G can coexist with EPC and LTE.This decomposition contrasts with the combined service functions of legacy EPC gateways.
  • NG Core Functions: The NG Core supports network slicing, NFV, service-function chaining, and SDN to improve core-network scalability, flexibility, and management.AMF handles access and mobility, SMF manages data paths, UPF defines data-path characteristics, and PCF manages policy and resources.
  • Deployment Considerations: 5G backhaul designs can reduce control-plane signaling and support ULL flow setup, but deployment choices require evaluation of latency, overhead, interfaces, distances, and cost.NG Core deployments may improve backhaul management while requiring high expenditures, and control-plane separation requires careful overhead consideration.
  • Wireless Access: Shorter TTIs and reduced processing time are the main wireless-access strategies for lowering 5G latency.Pipelining independently processed sub-blocks can reduce processing time, but adds resource-mapping and demapping overhead.
  • Wireless Trade-offs: Increasing OFDM subcarrier spacing reduces TTI duration but requires larger guard bands, tighter synchronization, sensitive receivers, and more hardware complexity.These requirements protect against inter-carrier and inter-symbol interference.
  • Wireless Scheduling: Proactive granting can reduce packet delay to less than 4 ms by eliminating scheduling-request and grant procedures.With a 1 ms LTE TTI, the request-and-grant RTT is at least 4 ms, producing data-transmission delays of 8 ms or more.

2) Fronthaul:

5G fronthaul designs balance stringent latency requirements against packetization, scheduling, bandwidth allocation, and transport constraints. Proposed optical, Ethernet, and predictive allocation techniques improve multiplexing or latency, but their benefits depend on traffic patterns and fronthaul split choices.

  • Fronthaul connects radio nodes, commonly RRUs, to radio processing nodes, commonly BBUs; CRAN centralizes and virtualizes BBU functions.
  • Optical Transport Techniques: 200 µs is the total fronthaul propagation-delay budget for a 1 ms 5G latency guideline, with processing targeted near 5 µs on a 20 km link.
  • Packetization and Scheduling over Ethernet: Scheduling shares Ethernet resources but can add queuing delay, while large frames increase waiting time and therefore favor short frames for lower latency.
  • Packetization and Scheduling over Ethernet: Packetization techniques combined with LRB scheduling and packet discarding produced significant multiplexing gains and increased the maximum RRHs supported on an Ethernet link.The evaluation varied look-ahead depth, prefetch, packet size, link capacity, scheduling policy, and functional split.
  • TDM-PON Dynamic Bandwidth Allocation: Polling DBA typically yields millisecond-scale end-to-end PON delay, exceeding fronthaul requirements of a few microseconds.The mechanism reports ONU demands and waits for centralized grants from the OLT scheduler.
  • TDM-PON Dynamic Bandwidth Allocation: 35 µs end-to-end fronthaul latency accompanied a 58% effective link-bandwidth-utilization increase over FBA under adaptive allocation.The allocation uses estimated long-term traffic statistics and is updated over successive time periods.

4) Network Management:

ULL network management must coordinate resource allocation, reliability, congestion control, and QoS across heterogeneous fronthaul, backhaul, and core paths. SDN offers integration and reconfiguration, but handovers, control-plane overhead, waveform timing, and traffic priorities remain important constraints.

  • Heterogeneous protocols across fronthaul and backhaul make comprehensive end-to-end management complex without an interoperable mechanism.
  • Optical Wireless Networking: A proposed optical-wireless 5G backhaul uses small cells, PONs, dedicated point-to-point links, and a dark-fiber protection ring to support capacity and latency requirements.
  • SDN Based Evolved Packet Core Networks: An SDN-based EPC replaces S-GW control-plane functions with an SDN controller and data-plane functions with SDN switching nodes, reducing large GTP messages to smaller control messages.The switching nodes can also assist attach and mobility procedures.
  • Temporary handover disruptions can cause buffer-bloat problems, while existing studies have not considered control-plane latency and complexity sufficiently.The discussed backhaul work focuses on DBA mechanisms for LTE X2 and S1 interfaces.
  • SDN manages fronthaul, backhaul, and core networks while reducing dedicated network-function entities and protocol overhead through dynamic packet-header manipulation.

4) Impact of Synchronization Inaccuracy:

Synchronization inaccuracies can disrupt time-triggered TSN scheduling, while broader ULL deployment also faces inter-networking, scalability, and resource-tradeoff challenges. The survey identifies these limitations across TSN, DetNet, and 5G-related mechanisms and calls for coordinated evaluation and future research.

  • Synchronization accuracy: Synchronization errors can shift time-aware transmissions into adjacent slots and cause short frames to miss their schedules.The resulting performance impact depends on traffic priorities, frame sizes, and timing-offset durations.
  • Network interoperability: TSN lacks an established mechanism for interoperating with external non-TSN networks, motivating centralized SDN or ingress-egress management.These approaches would identify external flows and configure reservations or flow paths to preserve delay-sensitive properties and end-to-end integrity.
  • Flow integrity: Packet replication improves reliability and can reduce latency through disjoint paths, but it increases effective bandwidth requirements.The required balance between replication degree, bandwidth, reliability, and latency remains a network-design constraint.
  • DetNet virtualization: DetNet relies on time-sensitive L2 support for deterministic L3 flows, although SDN combined with NFV is proposed to enable L2-independent adaptation.Packet replication and fragmentation are identified as mechanisms that do not require timing information.
  • Research directions: The survey identifies numerous ULL-mechanism gaps and emphasizes evaluating how mechanisms cooperate rather than studying them only in isolation.It argues that configuration and cooperation across mechanisms will be important for effective ULL services.
Loading 1803.07673v3…