Service Data Unit Transfer Rate

Description

SDU transfer rate is a QoS performance parameter that measures the throughput achieved by a network entity in terms of successfully transferred Service Data Units (SDUs). It is calculated as the total number of successfully transferred SDUs in a transfer sample divided by the input/output time for that sample, where the input/output time is defined as the larger of the input time or the output time for the sample. This definition is adopted from ITU-T Recommendation X.140, ensuring alignment with international QoS vocabulary. An SDU is the data unit passed by an upper protocol layer to a lower layer for transmission; for instance, a PDCP SDU is the IP packet received at the PDCP entity from the upper layer.

Unlike raw bit-rate throughput measured at a physical or link layer, SDU transfer rate is a service-level measure. It counts complete, successfully transferred SDUs — not partial segments, retransmitted fragments, or failed deliveries. This makes it an end-to-end, integrity-aware metric: an SDU that was dropped and never successfully delivered does not count toward the numerator. The transfer sample is a defined observation window during which incoming and outgoing SDUs are tracked; the 'input/output time' convention ensures that the denominator reflects the actual time window in which the transfers occurred, taking the longer of the two boundary observations. This avoids the subtle issue of samples that arrive over a short input window but exit over a longer output window, by conservatively using the larger value.

In 3GPP systems, SDU transfer rate can be applied to bearers in UMTS, EPS bearers in LTE, or QoS Flows in 5G. It is distinct from the Guaranteed Bit Rate (GBR) and Maximum Bit Rate (MBR) parameters that appear in QoS profiles: GBR/MBR are contractual or policed targets, while SDU transfer rate is a measured outcome. Operators can use SDU transfer rate to verify that a bearer is meeting its throughput commitments, to feed into monitoring systems for SLA enforcement, and to detect degradation in radio or transport conditions that reduces effective throughput. It can be measured at various reference points, such as a UE's PDCP entity, a NodeB/eNodeB/gNB PDCP layer, or a UPF's GTP-U endpoint.

To measure SDU transfer rate, the network must be able to identify SDU boundaries, count successful deliveries, and observe both input and output times. In 3GPP implementations, this is typically achieved by timestamping SDUs at ingress, instrumenting counters at egress, and matching confirmed deliveries. The concept has proven robust across generations: even as transport technologies moved from ATM-based UTRAN to all-IP EPC and then to service-based 5G core, the definition of SDU transfer rate remained stable, which preserves comparability of throughput measurements across the decades of 3GPP evolution.

For engineers and students, the key insight is that SDU transfer rate is a service-visible throughput metric. It complements SDU transfer delay and packet error loss rate to form a trio of QoS observables — delay, throughput, and loss — that together characterise the quality experienced by user data. Its ITU-T X.140 lineage makes it interoperable with non-3GPP definitions, and its persistence from Rel-4 to Rel-19 demonstrates its foundational role in the vocabulary of telecommunication QoS.

Purpose & Motivation

SDU transfer rate exists to provide a rigorous, service-level measure of throughput that reflects successfully delivered user data units, not merely raw bits pushed onto a medium. In early QoS frameworks, operators and standards bodies needed a way to state and verify how many complete SDUs a network actually delivered over a given window. The ITU-T X.140 definition provided exactly that, and 3GPP adopted it to ensure consistency with international practice.

The metric solves the problem that link-layer throughput can be misleading: a radio link can report high bit rates while SDUs are lost, fragmented, or retransmitted, resulting in poor user-visible throughput. By counting only successfully transferred SDUs and normalising by the observation window, SDU transfer rate aligns measured throughput with actual service delivery. This makes it useful for SLA verification, capacity planning, and detecting degradation in end-to-end data delivery.

Historically, the term was introduced in Rel-4 alongside SDU transfer delay, giving 3GPP a paired vocabulary for delay and throughput. Its wording has remained stable through every later release, from UMTS through LTE and 5G, which shows that the foundational problem it solves — measuring effective data delivery at the service level — has not changed even as the underlying radio and core architectures have been completely redesigned.