Number of Missing PDCP SDUs

Description

The Number of Missing PDCP SDUs (NMP) is a critical counter maintained by the Packet Data Convergence Protocol (PDCP) layer in both the User Equipment (UE) and the base station (eNodeB in LTE, gNB in NR). It specifically tracks Service Data Units (SDUs) that are detected as missing from the expected sequence at the receiving PDCP entity. An SDU is considered 'missing' when the receiver receives a PDCP Protocol Data Unit (PDU) with a Sequence Number (SN) that is higher than the next expected in-sequence SN, indicating a gap in the sequence. This gap signifies one or more SDUs were lost, likely due to errors in the lower layers (RLC, MAC, PHY) or over the air interface.

The NMP counter is incremented for each detected missing SDU. Its operation is tightly linked to the PDCP's in-order delivery and reordering functions. For each bearer, the receiving PDCP entity (either uplink at the gNB or downlink at the UE) maintains a receive window and a 'Next_PDCP_RX_SN' variable. When a PDU arrives with an SN (X) greater than 'Next_PDCP_RX_SN', it implies all SDUs with SNs between 'Next_PDCP_RX_SN' and X-1 are missing. The NMP counter is increased accordingly. These missing SDUs may be recovered if the lower RLC layer operates in Acknowledged Mode (AM) and successfully requests retransmissions, but the PDCP NMP metric captures the initial loss event from its perspective.

NMP is a vital performance measurement for the network operator. It is often reported via performance management counters (PM counters) from the RAN to the management system. A high or rapidly increasing NMP value on a specific cell or for a specific UE is a strong indicator of poor radio conditions, congestion, or hardware issues. It directly correlates to user-perceived packet loss and degradation of services like VoIP or video streaming. This metric is used in algorithms for radio link monitoring, handover triggering (if loss is persistent), and QoS adjustment. In advanced implementations, it can feed into Machine Learning models for predictive cell outage detection or network optimization.

Purpose & Motivation

The NMP metric was introduced to provide a clear, standardized measure of packet loss at the PDCP layer, which is the last protocol layer before user-plane data is handed over to the core network (or to the application in the UE). Prior to its explicit definition, packet loss could only be inferred indirectly from lower-layer metrics (like RLC retransmissions or HARQ NACKs) or from end-to-end application metrics. These indirect measures could be ambiguous; for example, high RLC retransmissions don't necessarily mean packets were ultimately lost if they were recovered.

NMP solves the problem of accurately quantifying service data loss that is perceptible to end-user applications. It provides a direct, unambiguous count of data units that failed to arrive in sequence at the PDCP receiver. This is crucial for monitoring Service Level Agreements (SLAs), especially for services with strict loss requirements like voice (VoLTE) or real-time gaming. It allows operators to pinpoint whether packet loss is occurring in the radio segment (high NMP) or elsewhere in the end-to-end path.

Its creation was motivated by the need for more granular and layer-specific performance indicators in LTE and the move towards Self-Organizing Networks (SON). As networks became more complex and automated, precise metrics like NMP became essential for automated root-cause analysis and optimization. By monitoring NMP trends, the network can automatically trigger corrective actions, such as adjusting handover parameters, modifying antenna tilt, or even flagging a potential cell site failure, thereby improving overall network reliability and user experience.