EN-DC SRB3 Demystified

3GPP 37.340 says that it is up the secondary node to establish "SRB3", but what exactly does this mean and how is it done?

Simple answer: The establishment of a signaling radio bearer (SRB) 3 in EN-DC mode means that RRC Measurement Reports for NR quality can be sent directly to the SgNB. This enables the 5G node to make intra-SgNB handover decisions and start the handover execution without involving the master eNodeB of the connection.

To prevent confusion the figure below shows a simplified scenario in which the Complete/Acknowledgement messages are not mentioned although they will be seen in the message flow.

A prerequisite is the successful addition of 5G radio resources as described in an earlier blog post. After this is completed the UE in the example transmits user plane information over the NR cell with the physical cell ID (PCI) = 12. In the transport network this cell is identified by NR CGI = xxxx52 (where „xxxx“ stands for a valid PLMN-ID and gNodeB-ID).

In the figure below the SgNB sends a X2AP SgNB Modification Required message that carries an embedded NR RRC cG-Config message. This cG-Config message is transparently forwarded by the MeNB to the UE. When arriving at the UE it activates CSI reference signal measurements on the 5G frequency including the serving 5G cell as well as its neighbors. It shall be noticed that here the concept of the Special Cell (SpCell) applies as it was defined for LTE-A CoMP scenarios. 

Instead of the X2AP SgNB Modification Required message the information for activating the CSI reference signal measurements can alternatively transported using the X2AP SgNB Addition Request Acknowledge or X2AP SgNB Change Required message.

In step 2 the UE sends a NR RRC (3GPP 38.331) Measurement Report that indicates a stronger 5G cell (the neighbor cell with PCI = 11) was measured. It might be a vendor-specific implementation to send this NR RRC Measurement Report simultaneously over uplink channels of the LTE radio link where it is carried by the LTE RRC Uplink Information Transfer MRDC (Multi-RAT Dual Connectivity) as well as over NR radio links where it is forwarded by the SgNB to the MeNB embedded in a X2AP RRC Transfer message.

Indeed, it is the SgNB that makes the handover decision, but since the MeNB is in charge of the signaling connection the handover command (here: another NR RRC cG-Config message that orders to switch the 5G radio link to the cell with PCI = 11) must be transmitted to the MeNB by using another X2AP SgNB Modification Required message.

After the UE received the NR CC cG-Config message sent by the SgNB the HO is executed and the 5G cell with PCI = 11 becomes the new primary secondary cell of the EN-DC connection.

Figure: Measurement Configuration, Reporting and Execution for intra-SgNB  Handover 

5G Call Drops in EN-DC: A Thread for Service Quality?

As explained in the post about EN-DC setup the addition of 5G NR radio resources to an ongoing LTE connection provides additional bandwidth for user plane data transmission. And it seems to be fair to say that at least in social media today 5G speed test results, especially throughput measurements, are treated as the benchmark for EN-DC service performance. Hence, it is also logical that a loss of the physical 5G radio link (5G drop) could have a serious impact on user experience.

I write "could", because as a matter of fact many 5G drops will not be recognized by subscribers using non-realtime services including HTTP streaming.

Due to the dual connectivity of LTE Master eNodeB (MeNB) and Secondary gNodeB (SgNB) the signaling trigger points indicating a 5G drop are also a bit more complex compared to what we know from LTE. Indeed, both network nodes are able to release 5G radio resources abnormally using three different X2AP message flow scenarios as shown in figure 1.

Figure 1: Three Basic Signaling Flows for Abnormal Release of 5G Radio Resources

Which of these individual message flows will be found in the trace data depends on which of the two base stations is the first one that detects a problem on the 5G radio link.

A particular case that is seen quite often in live networks is illustrated in figure 2.

Figure 2: 5G Drop due to SGC Failure in UE

Here the trigger is a LTE RRC SCG Failure Information NR message sent by the UE to the MeNB. Thus, the MeNB requests the release of 5G radio resources, which is acknowledged and executed by the SgNB.

In addition (not show in the figures) also the GTP/IP-Tunnel for user plane transport between S-GW and gNB is released by the MeNB after successful completion of the X2AP SgNB Release procedure.

For the UE the 5G drop is not as serious as a drop of the LTE radio connection would be. It is just a fallback on plain LTE, so to say. And after the switching the GTP/IP-Tunnel back to a downlink endpoint at the eNB 4G payload transmission continues.

The longer the overall duration of the radio connection the higher is the risk that the 5G radio resources are lost during an EN-DC call. One of my favorite cases is a subscriber with a radio connection that last a bit more than two and a half hours - see figure 3.

Figure 3: Location Session Record of a Single Subscriber indicating a total number 340 SgNB Drops over 2:33 Hours

Thanks to the smart algorithms of NETSCOUT's TrueCall geolocation engine there is high confidence that she or he sits in an indoor environment, but is served by an outdoor 5G cell. Thus, the penetration loss of the 5G signal is significant. Due to the higher frequency the path loss has also higher impact on the 5G than on the 4G radio signal. This seems to be the main reason why the 5G radio link drops as often as 340 times, which leads to an overall 5G (SgNB) Drop Rate of 83% for this connection.

However, the impact on the subscriber experience might not be a serious one as a different KPI, the 5G EN-DC Duration Rate indicates. According to the Duration Rate 99.99% of all the time 5G radio resources have been available for the subscriber. This is possible, because as also shown in figure 2 within a relatively short time new 5G radio resources are allocated again to this connection. Even if the subscriber is watching e.g. a Netflix video the buffering of already downloaded data on the end user device should be sufficient to conceal the short interruption of the data transfer over 5G resources.

With rising amount of EN-DC traffic it might be rather problematic for the network to handle the additional signaling load originating from the frequent 5G additions and releases. In extreme cases this may even lead to congestion due to CPU overload in RAN nodes or virtual network functions.

For realtime services like Voice over New Radio (VoNR) the entire situation changes. Here even short interruptions of the user plane radio transmission can be perceived by subscribers so that the above discussed 5G Duration Rate KPI will become insufficient to estimate the service quality. Hence, this will drive the demand for a fully integrated view of 5G RAN and Core KPIs covering both, signaling and application quality. 

How the Addition of 5G Radio Resources Increases the Complexity of LTE Signaling Procedures

While everybody is excited about the growing number of 5G deployments and speed test results it is easy to forget that a highly reliable LTE core and radio access network is the prerequisite for 5G non-standalone (NSA) data transmission.

Indeed, the 5G radio resources are just added to the ongoing LTE connection to provide higher bandwidth that enables in turn higher throughput. In other words: the current 5G deployments are designed for and limited to the needs of enhanced Mobile Broadband (eMBB) traffic.

To boost the user experience a 4G and a 5G base station cooperate and bundle there joint resources in one radio connection. The whole scenario is known as E-UTRA-NR Dual Connectivity (EN-DC) and as a matter of fact this dual connectivity increases the complexity of the RAN signaling tremendously.

The figure below shows the two base stations involved in the radio connection. On the left side is the Master eNodeB (MeNB) that controls the entire signaling connection. On the right side sits the en-gNB, also called Secondary gNodeB (SgNB). The inconsistency of acronyms originates from 3GPP specs. 3GPP 37.340 "E-UTRA and NR Multi-connectivity" can be seen as an umbrella document that originally coined "MeNB" and "SgNB". However, when standarizing more details these acronyms have been replaced with Master Node (MN) and Secondary Node (SN) and the latter is named "en-gNB" when used in EN-DC scenarios. (Sure this spec has a lot more terms to offer an is a must-read for every acroynm enthusiast.)

However, these naming conventions defined in 3GPP 37.340 have not made it into the protocol specs, especially not into 3GPP 36.423 "X2 Application Part" that names its message set for enabling EN-DC consequently "SgNB ...." - as also shown in the figure.

By the way the SgNB should also not be imagined as a single network element. On the 5G side often a virtual RAN architecture is already deployed. In such a VRAN a gNB central unit (CU) controls several gNB distributed units (DUs) and multiple remote radio heads (RRHs) including the 5G antennas can be connected to each DU.

5G Radio Resource Addition in EN-DC Mode

Before 5G radio resources can be added to the connection a LTE RRC connection and at least a default bearer for the user plane including its GTP/IP-Tunnel between S-GW and eNB must have been successfully established.

The trigger for adding 5G resources to this call is mostly an inter-RAT measurement event B1 (not shown in the figure). However, also blind addition of a 5G cells have been observed in some cases where the 5G cell coverage is expected to overlap exactly the footprint of the LTE master cell. 

All in all, there can be a 1:1 mappig between 4G and 5G cells when antennas are mounted very close to each other and pointing into the same direction. However, it is also possible that several 5G small cells (especially when using FR2 frequency bands) are deployed to cover the footprint of a 4G macro cell. 

The end-to-end signaling that adds 5G resources to the connection starts with the X2AP SgNB Addition Request message (1). It contains information about the active E-RABs of the connection, UE NR capabilities and often the singal strenght of the 5G cell as measured before is included as well. The message triggers allocation of 5G radio resources in the SgNB.

Similar to a X2 handover procedure the X2AP SgNB Addition Request Acknowledge message (2) is used to transport a NR RRC CG-Config message (3) back to the MeNB where it is "translated" into NR RRC Connection Reconfiguration and NR RRC Radio Bearer Config messages that are sent to the UE enclosed in a LTE RRC Connection Reconfiguration message. In these messages beside the Cell Group ID the 5G PCI and the absolute SSB frequency (a synonym for NR ARFCN) are found. Both, 5G PCI and SSB frequency in combination represent the identity of a 5G cell "visible" for the UE on the physical 5G radio interface. 

To keep the figure more simple I have spared the "translation" process in MeNB and show instead as next step the combined LTE/NR RRC Connection Reconfiguration Complete (4) that is send by the UE back to the MeNB to confim activation of the 5G radio link. 

After this the UE and the SgNB are ready to the 5G resources for radio transmission. However, one important component is still missing: a new GTP/IP-Tunnel for transporting the payload from the core network's serving gateway (S-GW) to the SgNB. 

The gNB downlink transport layer address (gNB DL TLA) and its appropriate GTP Tunnel Endpoint Identifier (TEID) have been already to the MeNB in step (2). Indeed, there are some more TLAs and TEIDs found in this X2AP message, especially for data forwarding across the X2 user plane interface (not shown in figure).

The MeNB forwards the gNB DL TLA/TEID to the MME (6) where it is forwarded to the S-GW using GTP-C signaling in case the two core network elements are connected over S11 reference point. The uplink TLA/TEID on the S-GW side remain the same as assigned before during establishement of the E-RAB (not shown in figure). So the new tunnel is now ready to be used (7) and transmission of payload packet starts immediately. 

In step (8) the MME confirms the successful tunnel establishment to the MeNB.

To total duration of the entire procedure from step (1) to (8) sums up to slightly more than 100 ms under lab conditions and typically around 300 ms in the live network. 

This delay does not have a direct impact on user plane latency in the initial 5G setup phase. However, the subscriber experience might be different when it comes to inter-MeNB handover, because there is no direct handover between 5G neighbor cells. 

Changing the MeNB due to subscriber mobility means: release all 5G resources on the source (M)eNB side, perform intra-LTE handover to the target (M)eNB and add new 5G resources after handover is successfully completed.