EPS Bearers (Default & Dedicated) in LTE 4G

Introduction

The EPS bearer is the object around which the entire LTE user plane is organised. It is a logical connection with a QoS contract attached, running end to end from the UE to the P-GW (PDN Gateway), and it is the finest granularity at which LTE differentiates traffic. Every user-plane concept you meet in LTE — QoS (QCI/ARP), packet steering (TFT), core tunnelling (GTP-U/TEID), and the radio protocol stack (DRB/PDCP/RLC) — hangs off a bearer.

Bearers appear at specific moments in the UE lifecycle. The very first one, the default bearer, is created during attach and lives for as long as the UE keeps its PDN connection. Additional dedicated bearers are created on demand later — typically when a service such as a VoLTE call needs guaranteed treatment — and torn down when the service ends. Reading any LTE trace means tracking these bearers as they are set up, modified and released.

It matters because bearers are where policy becomes reality: an operator's decision that "voice must have bounded delay" turns into a GBR dedicated bearer at QCI 1 with a TFT that steers the RTP flow into it. Understand the bearer and you understand how a QoS intent becomes a scheduled packet.

Why EPS bearers are needed

Concretely, LTE needs bearers because different traffic needs different handling and the network must apply that handling consistently at every hop. Voice needs low, bounded delay and can tolerate a little loss; a file download needs throughput and hates loss. Rather than inspect every packet's application at every node, LTE groups flows with the same needs into one bearer and gives that bearer a single QoS label. Each node along the path — eNB, S-GW, P-GW — then treats the whole bearer identically, with no per-packet decision to make.

What an EPS Bearer Is

Think of the EPS bearer as an end-to-end tunnel with a promise attached. The promise is QoS — a defined treatment (priority, guaranteed rate or not, tolerable delay and packet loss) that every packet in that bearer receives. The tunnel runs between the UE and the P-GW (PDN Gateway), which sits at the edge of the operator's network and anchors the connection to the external packet data network — the internet, or an operator service network such as an IMS APN.

Beyond the P-GW there is no bearer — just an ordinary external IP connection to the internet or an application server. The bearer concept lives entirely inside EPS. That is the first mental boundary to draw: the EPS bearer is an operator-internal construct, and the outside world never sees it.

PDN Connection vs Bearer

These two terms are constantly confused, so pin them down before going further. A PDN connection is the association between the UE and one external packet data network (one APN), and it is what gives the UE an IP address. A bearer is a QoS pipe inside a PDN connection. One PDN connection therefore contains one or more bearers: exactly one default bearer plus zero or more dedicated bearers.

A UE can hold several PDN connections at once — for example one to an internet APN and one to an ims APN — and each is a separate IP context with its own IP address and its own default bearer. So the hierarchy reads: UE → one or more PDN connections → each with one default bearer and optional dedicated bearers → each bearer carrying one or more IP flows. The APN selects which external network you reach; the bearer decides how well your packets are treated on the way there.

The Bearer as Concatenated Segments

The end-to-end EPS bearer is not one physical connection; it is a chain of three segments stitched together, each hop bound to the next so the QoS treatment carries through. The core-network segments are realised as GTP-U tunnels — user-plane GPRS Tunnelling Protocol (TS 29.281), carried over UDP port 2152, where each tunnel endpoint is named by a TEID (Tunnel Endpoint Identifier), a 32-bit value that the peer stamps into the GTP header of every packet.

Concatenated: DRB + S1 bearer + S5/S8 bearer = the EPS bearer. The eNB binds a DRB to an S1 bearer, and the S-GW binds that S1 bearer to an S5/S8 bearer, so a single IP packet passes through all three segments with the same QoS treatment end to end. The S-GW is a pure relay in the user plane: it does no QoS reclassification, it simply swaps one GTP-U tunnel for the next by mapping the incoming TEID to the outgoing one. Beyond the P-GW lies only the external IP connection to the PDN, with no bearer at all.

The Default Bearer — always on

When a UE attaches to LTE, the network immediately establishes one bearer for it: the default bearer. It is created during the attach procedure and stays up for as long as the UE keeps its PDN connection to that APN — you get exactly one default bearer per PDN connection, and it never goes away while the connection exists. Its job is to keep the pipe alive and give the UE always-on IP connectivity from the moment it attaches.

A default bearer is always non-GBR (no guaranteed bit rate). It is best-effort: the network reserves no fixed rate for it, so it flexes with available capacity — ideal for general internet traffic. It is assigned a QoS profile (a QCI and an ARP) at setup and keeps that profile for its whole lifetime. A typical internet default bearer runs at QCI 9 (or 8/6), while an IMS signalling default bearer runs at QCI 5.

The Dedicated Bearer — QoS on demand

A dedicated bearer is an additional bearer created on top of an existing default bearer to give a specific set of flows their own, usually stronger, QoS. It is always linked to a default bearer — it shares that PDN connection and its IP address — and it is set up on demand, typically triggered by the PCRF pushing a policy (a PCC rule) to the P-GW when an application such as an IMS voice session needs guaranteed treatment. The link back to its default bearer is carried explicitly in the Linked EPS Bearer Identity (LBI), which names the default bearer's EBI.

Unlike the default bearer, a dedicated bearer can be either GBR or non-GBR. The classic example is VoLTE: when a voice call is set up, IMS asks the PCRF for a GBR dedicated bearer at QCI 1, which reserves bandwidth so the voice media gets bounded delay and low loss for the duration of the call. When the call ends, the dedicated bearer is released and the default bearer stays. A non-GBR dedicated bearer is also possible — for example to give a particular application better priority (a lower QCI) without reserving a fixed rate.

Traffic Flow Templates — steering packets to the right bearer

Once a UE has more than one bearer, something has to decide which packet goes on which bearer. That job belongs to the Traffic Flow Template (TFT) — a set of packet filters that map IP flows to a particular bearer. Each filter matches on the classic IP 5-tuple: source IP address, destination IP address, source port, destination port, and protocol (plus optionally the ToS/DSCP and IPsec SPI). A packet that matches a dedicated bearer's filter rides that bearer; anything that matches nothing falls through to the default bearer.

Each packet filter within a TFT carries a packet filter identifier, an evaluation precedence value (0–255, lower is evaluated first), and the filter components themselves (each 5-tuple element is a distinct component type). Precedence matters because filters are tried in order and the first match wins, so a more specific filter must be given a lower precedence value than a broad catch-all. So a VoLTE flow's RTP media, matched by its 5-tuple, is steered onto the QCI 1 dedicated bearer, while ordinary web traffic falls through to the default bearer. The default bearer usually has a match-everything filter (or no explicit TFT at all), so anything not claimed by a dedicated bearer's TFT rides the default bearer — there is always a home for every packet. The P-GW installs both directions when it creates or modifies a bearer: it keeps the DL TFT locally and sends the UL TFT down to the UE (inside the bearer-setup NAS signalling), so both ends classify the same flows onto the same bearer consistently.

The Bearer Identifiers

A single EPS bearer is referred to by several different identifiers, one per interface, because each node names the bearer in its own context. Keeping them straight is essential when reading traces: the same bearer appears as an EBI in NAS, an E-RAB ID in S1AP, and a DRB ID in RRC.

The neat part is the numeric identity between EBI and E-RAB ID: the same value threads NAS and S1AP, so a bearer set up as EBI 6 in the attach signalling appears as E-RAB ID 6 when the MME asks the eNB to set up the radio side. The eNB then chooses a DRB ID for the air interface and maps it to that E-RAB. On dedicated bearers, the LBI is what stops the network from losing track of which default bearer — and therefore which PDN connection and IP address — a dedicated bearer hangs off.

Mapping onto the Radio — DRB to Logical Channel

On the radio side, the eNB maps each EPS bearer to a Data Radio Bearer (DRB) — the radio segment of the chain — and the DRB runs down the LTE user-plane protocol stack. There is exactly one DRB per EPS bearer, so the QoS distinction you set up in the core survives all the way down to how the MAC scheduler prioritises the corresponding logical channel.

So the mapping is a clean chain: an IP flow → classified by TFT onto an EPS bearer (named by its EBI) → carried on a DRB over the air → served by a dedicated PDCP and RLC entity → a DTCH logical channel that MAC schedules onto the shared transport channel. Because there is one DRB per EPS bearer and one logical channel per DRB, QoS is not re-derived at each layer — it is inherited straight down the stack from the bearer it belongs to.

Default vs Dedicated — side by side

The two bearer types differ in when they are created, what QoS they can carry, and their relationship to the PDN connection. This is the table interviewers reach for most often:

Summary

The EPS bearer is the atom of LTE connectivity: one end-to-end pipe from the UE to the P-GW, carrying one agreed QoS. Draw the boundaries in this order and the whole user plane falls into place — a PDN connection (one APN, one IP address) contains exactly one default bearer (always-on, non-GBR, born at attach) plus zero or more dedicated bearers (on-demand, GBR or non-GBR, linked back by LBI).

End to end, the bearer is really three concatenated segments — DRB + S1 bearer + S5/S8 bearer — with the two core segments realised as GTP-U tunnels named by TEIDs. Packets are steered onto the right bearer by TFT filters at the two edges (UE for uplink, P-GW for downlink), and the same bearer is named differently at each interface: EBI in NAS, E-RAB ID in S1AP (numerically equal), DRB ID in RRC. On the radio it is one DRB per bearer, one logical channel per DRB, so QoS is inherited straight down the stack.

Hold onto the one principle that everything else follows from: one EPS bearer = one QoS. A new QoS requirement means a new bearer — and that rigidity is exactly what 5G's per-flow (QFI) model later relaxes.