featured 38859

LTE Transport Network: Key Concepts, Architecture, and Common Issues

1.    Introduction

While 3GPP specifies the Radio Network Layer (RNL) for LTE, the Transport Network Layer (TNL) relies on standards from outside 3GPP. The main standardization bodies that define TNL technologies include IETF (Internet Engineering Task Force), IEEE (Institute of Electrical and Electronics Engineers), ITU-T (International Telecommunication Union — Telecommunication Standardization Sector) and MEF (Metro Ethernet Forum).

This document provides an overview of common LTE transport concepts, requirements and typical deployment considerations.

2.    LTE Network Architecture

This section does not attempt a full description of LTE architecture; many comprehensive tutorials exist. Key characteristics of LTE networks include:

  • Flat architecture reducing hierarchical layers
  • Elimination of the circuit-switched (CS) domain, moving to an all-packet core
  • Main LTE network elements:
    • Mobility Management Entity (MME): manages mobility and maintains UE context
    • SAE Gateway (SAE-GW): logically includes the Serving Gateway (S-GW) and the Packet Data Network Gateway (P-GW)
    • eNodeB: provides the radio access (air interface) to the UE

3.    LTE traffic types

The Evolved Packet System (EPS) defines several reference points used in LTE transport:

X2 is the reference point carrying control and user plane traffic between two eNodeBs.

S1-U is the user plane reference point between eNodeB and the SAE-GW.

S1-C is the control plane reference point between eNodeB and the MME.

The synchronization plane (S-plane) connects the eNodeB synchronization slave to the network synchronization grandmaster.

The management plane (M-plane) is not specified in detail by 3GPP but is used for O&M traffic.

4.    LTE Transport Requirements

4.1. Throughput (Capacity)

LTE supports high peak cell rates thanks to wide radio bandwidths. For example, a 20 MHz channel can deliver peak downlink rates around 150 Mbps (64QAM, 2×2 MIMO) and peak uplink rates up to 75 Mbps (64QAM, single stream) under ideal conditions (close to the base station, minimal interference).

Average cell capacity is also important: LTE typically achieves higher spectral efficiency than previous systems. Simulations for LTE Release 8 indicate typical spectral efficiency values (e.g., around 1.75 b/s/Hz downlink and 0.75 b/s/Hz uplink under certain conditions).

Transport engineering tasks include:

  • Dimensioning last-mile bandwidth for a single eNodeB
  • Dimensioning aggregate bandwidth for multiple eNodeBs

Transport capacity can be estimated in two ways:

  • From air interface capabilities (quick, useful for initial planning)
  • From operator-specific user traffic profiles (requires detailed traffic data)

4.1.1    Last-mile bandwidth dimensioning

X2 traffic

X2 is mainly used for user traffic signaling during UE handover between eNodeBs. X2 traffic is commonly expressed as a fraction of S1 traffic. NGMN recommends a typical X2 overhead of about 4% of S1 traffic; sites with high user mobility may require up to 10%.

Control Plane, OAM and Synchronization Signaling

Control plane signaling on S1 and X2 is small compared to user plane traffic (on the order of ~0.3 Mbps). M-plane and synchronization signaling (S-plane) also consume modest capacity (~1 Mbps or less in many cases).

User Plane

To convert air-interface capacity into transport capacity, subtract radio protocol overhead and then add transport protocol overhead.

On the radio side, PDCP, RLC and MAC protocol headers average roughly 9 bytes; with a realistic packet-size mix this corresponds to about 2% air-interface overhead.

Transport-layer overhead on S1-U and X2-U depends on whether IPsec is used. Typical header totals are:

  • With IPsec: GTP-U + UDP + IP + Ethernet + IPsec related headers ≈ 144 bytes
  • Without IPsec: GTP-U + UDP + IP + Ethernet ≈ 78 bytes

For a traffic mix of 50% small (60 bytes), 25% medium (600 bytes) and 25% large (1500 bytes) IP packets, transport overhead is approximately 27% with IPsec and 15% without IPsec.

Transport Overhead Number of bytes
GTP-U 8 bytes
UDP 8 bytes
IPv4 Header 20 bytes
(optional) IPsec ESP Header 8 bytes
(optional) IPsec AES IV 16 bytes
(optional) IPsec ESP Trailer 10 bytes
(optional) IPsec Authentication (ICV) 12 bytes
(optional) IPsec Tunnel IPv4 Header 20 bytes
Ethernet Header (incl. 4 bytes VLAN) 22 bytes
Ethernet IF Gap, Preamble & SFD 20 bytes
Total 144 bytes (with IPsec) / 78 bytes (without IPsec)

Subtracting radio overhead (~2%), the transport capacity addition becomes roughly +25% with IPsec and +13% without IPsec over the air-interface data rate.

Dimensioning Based on Cell Air Interface Capacity

Three dimensioning approaches are common. Let N be the number of cells at an eNodeB site, Cpeak and Cavrg the peak and average capacity of one cell, and Ctrs the required transport capacity.

  • All-Average: Ctrs = N × Cavrg — transport supports aggregate average traffic.
  • All-Average/Single-Peak: Ctrs = max(N × Cavrg, Cpeak) — supports aggregate average, and at least the peak capacity of one cell.
  • All-Peak: Ctrs = N × Cpeak — supports aggregate peak of all cells (generally over-dimensioned and costly).

All-Average/Single-Peak is often a pragmatic compromise, but may under-provision hot-spots or over-provision lightly used sites.

All-Average/Single-Peak last-mile example

For an eNodeB with three sectors, 10 MHz channels, downlink support 64QAM 2×2 MIMO and uplink 64QAM single stream, a practical choice is:

Downlink Ctrs = max(3 × average_cell_DL, peak_cell_DL) = 75 Mbps (peak-limited)

Uplink Ctrs = max(3 × average_cell_UL, peak_cell_UL) = 37.5 Mbps (peak-limited)

Applying overhead: Downlink transport capacity with IPsec ≈ 75 Mbps × 1.25 ≈ 94 Mbps; Uplink transport capacity with IPsec ≈ 37.5 Mbps × 1.25 ≈ 47 Mbps.

4.1.2    Aggregate link bandwidth dimensioning

For aggregation links, a common estimate is:

Transport Capacity = #cells × average cell capacity + transport overhead

Example: average cell capacity = spectral efficiency × channel BW = 1.75 × 10 MHz = 17.5 Mbps; transport overhead typically ~25% with IPsec.

For a network of six eNodeBs (3 cells each), this approach lets you estimate the required aggregate transport capacity for the uplinks and downlinks toward aggregation/core.

4.2. Delay and Delay variation

Delay and packet delay variation requirements differ across planes: user plane (U-plane), control plane (C-plane), and synchronization plane (S-plane).

4.2.3    U-plane

NGMN recommends that the backhaul should guarantee an end-to-end maximum two-way delay of 10 ms where required, and in general aim for 5 ms where the operator demands tighter performance.

Delay influences transport layer performance, notably TCP throughput and user experience.

TCP and RTT

A single TCP connection’s steady-state throughput is limited by the TCP window size divided by RTT. With a 64 KB TCP receive window, a 20 ms RTT caps a single TCP stream around 25 Mbps. Browsers mitigate this with multiple parallel TCP connections and TCP Window Scaling, but initial transfer phases (TCP slow start) are still affected by RTT. Because typical web page transfers often complete during slow start, RTT remains an important factor for perceived web performance.

4.2.4    C/M-plane

Transport latency requirements for the control and management planes are generally less stringent than for critical user-plane services, but overall LTE transport latency is driven primarily by U-plane needs.

4.2.5    S-plane

Legacy 2G/3G base stations often obtained synchronization from TDM networks or GPS. LTE moves to IP-based transport for backhaul, requiring packet-based synchronization methods such as IEEE 1588 Precision Time Protocol (PTP) and Synchronous Ethernet (SyncE).

PTP requires low jitter for synchronization packets; robust PTP slave implementations can tolerate jitter up to approximately ±5 ms for frequency synchronization. For LTE TDD, which requires tight phase alignment (e.g., on the order of ±1.5 μs), jitter and delay asymmetry must be strictly controlled. Intermediate network nodes handling synchronization traffic should support boundary clock or transparent clock functionality to preserve timing accuracy.

SyncE (per ITU-T G.8261/G.8262/G.8264) distributes frequency at Layer 1 and provides stable recovered frequency regardless of packet load, but all intermediate nodes in the synchronization path must support SyncE to be effective.

5.    LTE Transport Network Architecture

A typical mobile backhaul structure includes access, aggregation and core domains. Different architecture options are adopted depending on operator preferences and transport technologies available.

Common architectures include:

  • Carrier Ethernet: Pure IEEE 802.1ad-based solution using service VLANs (S-VLAN) to carry customer VLANs (C-VLAN) across the Ethernet domain. One or more MEF EVCs can span access and aggregation.
  • Carrier Ethernet + L2/L3 VPN: Access uses Carrier Ethernet while aggregation uses MPLS; MEF EVCs can span the access domain with L2/L3 VPNs in aggregation.
  • MPLS access + L2/L3 VPN: MPLS (including MPLS-TP) is used in access and aggregation to create connection-oriented flows and to enter VPNs into aggregation.
  • L2/L3 VPN in access + L2/L3 VPN in aggregation: Typical combination is L2 VPN in access and L3 VPN (IP/MPLS) in aggregation—MEF EVCs are not used in this case.
  • End-to-end (multi-segment) pseudowire: Uses pseudowire circuits for access and aggregation; MEF EVCs are typically not used.
  • Full L3: MPLS/MPLS-TP from the eNodeB through to MME and S-/P-GW controllers. Both access and aggregation belong to one logical MPLS domain; VPNs may be L2 or L3.

5.1. Carrier Ethernet deployment scenarios

MEF defines Ethernet service types used in Carrier Ethernet networks, notably E-Line, E-LAN and E-Tree.

5.1.1    E-Line Service Type

E-Line provides point-to-point Ethernet Virtual Connections (EVCs) between an eNodeB and an edge router or security gateway. Each eNodeB can be associated with a specific VLAN ID. The edge router handles routing of S1 traffic between eNodeBs and the core, as well as inter-group X2 routing when needed.

5.1.2    E-LAN Service Type

E-LAN creates a multi-point LAN segment for a group of eNodeBs, typically identified by a single VLAN ID. Within the E-LAN, eNodeBs are distinguished by MAC address. S1 and X2 traffic among eNodeBs in the same group can be switched locally inside the E-LAN, while S1 to the core and X2 between groups is routed by the edge router.

A drawback of E-LAN is that it forms an open broadcast domain—layer 2 broadcasts reach all leaf nodes—so administrators must consider broadcast containment. E-LAN is often recommended when IPsec is not required.

5.1.3    E-Tree Service Type

E-Tree provides a rooted multipoint service where the edge router acts as root and direct connectivity between leaf nodes (eNodeBs) is not permitted. E-Tree prevents L2 broadcast propagation between leaf nodes, reducing certain attack vectors. If IPsec is not used, E-Tree may introduce greater configuration complexity; in such cases, E-LAN can be a simpler option.

6.    LTE Transport QoS

3GPP defines QoS at the RAN and core service layers, but it does not mandate mobile backhaul transport-layer QoS. Aligning transport-layer QoS with LTE service-layer QoS is therefore essential. Industry groups such as IETF, MEF and the Broadband Forum work on defining transport QoS models suitable for mobile backhaul.

MEF’s class-of-service models and NGMN’s service flow classification provide practical guidance for mapping LTE service requirements to transport-layer QoS.

NGMN’s service flow classification survey revealed variation among operators in how many CoS levels they use (typically 4–8) and the priorities assigned. Some treatments are consistent: QCI 9 often receives the lowest priority and packet synchronization receives the highest across respondents. Management plane treatment varied widely, likely due to differing definitions of management traffic in respondents’ networks.

6.1. NGMN Service Flow Classification Proposal

NGMN proposes multiple CoS schemes to suit different operator needs:

  • Two-CoS scheme
  • Three-CoS scheme
  • Four-CoS scheme

6.1.1    Two-CoS Classification Scheme

C1 Voice, Real-Time Gaming, Synchronization and Control Plane/OAM
C8 Everything else

Target performance indicators for this scheme set strict requirements for C1 (for example, packet loss targets around 10^-6 and low delay/very low jitter), while C8 carries best-effort traffic with no specific guarantees at the transport layer.

3GPP suggests that a representative transport delay between the PCEF (Policy and Charging Enforcement Function) and the base station of around 20 ms may be acceptable for many applications, though tighter bounds are required for some services.

6.1.2    Three-CoS Classification Scheme

C1 Voice, Real-Time Gaming, Synchronization and Control Plane/OAM
C2 2G Data (EDGE) and Real-Time Video
C8 Everything else

This scheme defines tighter performance targets for C1 and C2 (e.g., similar loss objectives for both classes, but larger delay and jitter targets for C2 compared to C1).

6.1.3    Four-CoS Classification Scheme

C1 Voice, Real-Time Gaming, Synchronization and Control Plane/OAM
C2 2G Data (EDGE) and Real-Time Video
C3 Premium Data (buffered video, non-GBR real-time)
C8 Everything else

The four-class scheme adds a middle priority class (C3) for premium non-real-time traffic. Target objectives tighten progressively from C8 (best-effort) up to C1 (highest priority and strict QoS).

6.2. Inter-layer Class of Service Alignment

Transport QoS must be aligned across layers: RAN service QoS maps to IP-layer QoS, which in turn maps to Ethernet/MAC-level CoS. A consistent inter-layer CoS alignment ensures that high-priority LTE services (e.g., synchronization, voice, control plane) receive appropriate queuing, scheduling and loss/delay guarantees across the entire transport path.

7.    References

[REF1] THE LTE/SAE DEPLOYMENT HANDBOOK – Edited by Jyrki T. J. Penttinen

[REF2] Guidelines for LTE Backhaul Traffic Estimation by NGMN Alliance

[REF3] Next Generation Mobile Networks Optimised Backhaul Requirements, NGMN

[REF4] Integrated QoS Management by NGMN Alliance

[REF5] MEF 22 – Mobile Backhaul Implementation Agreement

[REF6] LTE backhauling deployment scenarios by NGMN Alliance

[REF7] MEF 6.2 EVC Ethernet Services Definitions Phase 3