5G x-Haul design over SRv6 without link congestion

SRv6 is used in 5G networks to enable slicing and support virtual networks with diverse requirements, such as ultra-low latency and high reliability, on shared infrastructure. This is essential for applications like IoT and virtual reality.
In recognition of more and more networks evolving toward SRv6, we defined a time synchronization test case for a network with SRv6 as the transport protocol between O-RAN components (Open-Radio Unit (O-RU), Open-Central Unit (O-CU), and Open-Distributed Unit (O-DU)) in a multi-vendor environment.
These test cases validate that the SRv6 architecture does not significantly impact the time synchronization with and without link congestion, ensuring precise time synchronization and interoperability across the transport network in accordance with the requirements of "O-RAN WG9.XTRP-TST".

Figure 101

Figure 101: 5G x-Haul over SRv6 - setup


For this test, the Microchip TimeProvider® Telecom Grandmaster (T-GM) sent PTP and SyncE (no enhanced SyncE) downstream, with the upper path selected by the Best TimeTransmitter Clock Algorithm (BTCA) as the primary time synchronization path because it had one fewer hop/step removed compared to the lower path.
For the link congestion, background traffic was started and routed over the upper path using 100GE links; the link congestion happened between the upper Pre-Aggregation node and the two upper Access nodes.
Additionally, O-RAN Fronthaul traffic between the emulated O-DU and O-RU was injected into the upper Pre-Aggregation node and the upper two Access nodes, further congesting the link.

We observed that the whole network passed this test without link congestion between the Pre-Aggregation Node and the Access Nodes during testing. The accumulated time error between the Microchip TimeProvider® 4500 PRTC T-GM  and the output of the Access Nodes was less than +/—15ns, which is significantly less than the limit of 1100ns. 

Figure 102

Figure 102: Time Error Measurement by Keysight Time Sync Analyzer

Figure 103

Figure 103: Time Error Measurement by Calnex Paragon-Neo PAM4

The relative time error between the Access Nodes was also significantly less than the required +/—130ns at less than +/- 8ns.

Figure 104

Figure 104: Relative Time Error Measurement between Juniper ACX7024 and Huawei ATN 910D-A by Keysight Time Sync Analyzer

Figure 105

Figure 105:  Relative Time Error Measurement between Ericsson R6671 and H3C S12500R-48Y8C by Calnex Paragon-Neo PAM4

As both requirements were strongly met, this test was a pass.
      
For the link congestion test, four different tests were run; only one Pre-aggregation to the access node link was congested during each run, so this test was repeated for each access node.
After the links were congested, not much changed. Effectively no change in time error was observed compared to without link congestion, and the congested devices dropped no PTP packets, thus also resulting in a pass and proving that an SRv6 Architecture does not impact the timing performance, even if link congestion is present.