How it Works
The even distribution of ingress traffic across all the available worker threads on the UPF is ensured by creating the unique 3-tuple hash entries.
The GTPU header source IP is the gNodeB’s N3 service egress IP, while the source port is the standard port in the absence of port randomization support on the gNodeB. With these two values being constants per gNodeB, the only variable that can be achieved in the GTPU header is the destination IP (N3 service ingress IP address).
The following points describe how the unique 3-tuple hash entries are created:
-
GTPU addresses are assigned in FTEID in Sx Session Establish Response for a subscriber session in a round-robin manner. However, this round-robin mechanism is ensured only within the session manager.
-
When a session manager is selected, the GTPU address in FTEID is allocated in a round-robin manner to all the sessions landing on the selected session manager.
-
Deletion of IP address from the GTPU service or the interface does not cause the GTPU service to stop. The existing calls will be deleted. However, the uplink traffic of the subscriber to which the deleted IP address was assigned will not be received at UPF.
-
If every IP address is configured under a separate interface, the interface must be configured as a loopback (multiple non-loopback interfaces cannot share the same subnet). If any interface goes down and comes alive again, the GTPU service does not stop as the interfaces are configured as loopback.
-
Every address pair is one GTPU endpoint for a GTPU service. Every endpoint should belong to the same subnet.
Note |
|