5 Reference Architecture and Procedures

26.5653GPPRelease 18Split Rendering Media Service EnablerTS

Editor’s Note: The terminalogy and reference-point naming convention needs to be aligned in this document.

5.1 Reference Architecture

5.1.1 Introduction

We define different variants of the reference architecture for the split rendering MSE, each representing a different perspective and level of detail.

5.1.2 Client Architecture

Figure 5.1.2-1 depicts the client architecture.

Figure 5.1.2-1 – Device architecture of AR UE

5.1.3 End-to-End Architecture

The end-to-end architecture is depicted in figure 5.1.3-1.

Editor’s Note: this architecture variant is TBD based on the MeCAR discussions.

5.1.4 Control Architecture

Figure 5.1-3 – Split management architecture

5.1.5 User Plane Architecture

Figure 5.1.5-1 depicts the user plane architecture for split rendering.

5.2 Procedures and Call Flows

5.2.1 Call flow for Split Rendering instance discovery

5.2.1.1 Client-driven procedures and call flows

Figure 5.2.1-1 demonstrates a call flow for setting up the split rendering by the client.

Figure 5.2.1-1: High-level call flow for initiating a split

Steps:

  1. The 5GMS Application Service Provider requests the SRF the provisioning a split management session.
  2. The split management session is announced to the Application as part of the Service Access Information.
  3. The 5GMS-Aware application requests a split of the client media functions from the SRC.
  4. The SRC inquires the Media Service Client to discover the client’s media capabilities.
  5. The SRC requests the SRF split of the client media functions.
  6. The SRF inquires the 5GMS AS about its capabilities.
  7. The SRC and SRF negotiate on the acceptable capabilities for the device and agree on the split option.
  8. The SRF requests the 5GMS AS to start the split-rendering process on the edge.
  9. The SRF acknowledges the SRC that the split-rendering on edge is running and provides its access information.
  10. The SRC acknowledges the Application that the split-rendering on edge is running.
  11. The 5GMS-aware Application request starting a media session from the 5GMS client.
  12. The 5GMS client connects to the split-rendering on the 5GMS AS using provided information in step 9.

5.2.1.2 Network-driven procedures and call flows

Figure 5.2.1.2-1 demonstrates a call flow for setting up the split rendering by the client.

Figure 5.2.1.2-1: High-level call flow for the network-driven split management

Steps:

  1. The 5GMS Application Service Provider requests the SRF the provisioning a split management session.
  2. The split management session is announced to the Application as part of the Service Access Information.
  3. The SRF inquires the 5GMS AS about its capabilities.
  4. The SRF requests the SRC split of the client media functions.
  5. The SRC inquires the Media Service Client to discover the client’s media capabilities.
  6. The SRC and SRF negotiate on the acceptable capabilities for the device and agree on the split option.
  7. The SRF requests the 5GMS AS to start the split-rendering process on the edge.
  8. The SRF acknowledges the SRC that the split-rendering on edge is running and provides its access information.
  9. The SRC acknowledges the Application that the split-rendering on edge is running.
  10. The 5GMS-aware Application request starting a media session from the 5GMS client.
  11. The 5GMS client connects to the split-rendering on the 5GMS AS using provided information in step 9.

5.2.2 Call flow for Split Rendering session setup

The split rendering operation can be described by the following call flow:

The steps are:

  1. The Presentation Engine discovers the split rendering EAS and sets up a connection to it. It provides information about its rendering capabilities and the OpenXR configuration.
  2. In response, the split rendering EAS creates a description of the split rendering output and the input it expects to receive from the UE.
  3. The Presentation Engine requests the buffer streams from the MAF, which in turn establishes a connection to the split rendering EAS to stream pose and retrieve split rendering buffers.
  4. The Source Manager retrieves pose and user input from the XR runtime.
  5. The Source Manager shares the pose predictions and user input actions with the split rendering EAS
  6. The split rendering EAS uses that information to render the frame
  7. The rendered frame is encoded and streamed down to the MAF