ON THIS PAGE
Results Summary and Analysis
Within the course of validation, we were able to demonstrate a robust solution for 5G Fronthaul and Midhaul transport infrastructure utilizing Seamless MPLS. The JVD generates reasonable multi-vector scale of L2/L3 connectivity services as compared with MNO and MAN operator expectations for real network deployments, while satisfying stringent SLA requirements. The scale reference in Table 2 characterizes primary multidimensional KPI’s represented in the validated profile.
ACX7100-48L, ACX7100-32C and ACX7509 as the DUT routers have each successfully passed 186 test cases curated to support the given reference architecture.
With this network design, the architecture can deliver fast restoration within 50ms for most traffic flows transported over ISIS-SR with TI-LFA. Load distribution and optimization knobs were shown to improve service restoration against link/node failures. Link events consistently achieve <50ms convergence, see Table 2. Node failures were more disruptive as expected, exacerbated by scale and in some rare cases, production limitations. Contact your Juniper Networks representative for a complete results report and analysis.
ECMP Load Balancing
The design minimizes traffic issues when links or nodes fail by distributing the load across equal cost paths and making use of TI-LFA protection mechanisms. Multiple load-sharing methodologies are incorporated, such as adjusting IGP metrics, BGP multipath, ECMP fast-reroute, and VPN-unequal-cost for L3VPN services. We also activated FAT-PW label on ACX7000 devices where supported, including Layer 2 VPN and Layer 2 Circuits, which enables additional granularity to the hash computation. Hash-key configuration is mandatory on ACX7000 series to perform traffic load balancing. By default, there are no hash-keys enabled. This validation enables all relevant hash-keys for extracting L2, L3 and MPLS fields.
The full test report includes details about the load sharing configurations enabled for this validation. Junos OS Evolved 22.3R1-S1/R2 and later are recommended versions for ACX7000 series and includes additional load sharing enhancements. Contact your Juniper Networks representative if you’d like to receive the complete report.
ACX7000 ECMP Load Balancing Performance [1] | |||||||
---|---|---|---|---|---|---|---|
Service: DUT (Traffic Path) | ECMP Links | Flow | FAT-PW | Link1 LAG (#) | Link2 LAG (#) | Link3 LAG (#) | Link4 LAG (#) |
L2VPN: ACX7100-48L (AN3 to SAG) | 3 | 100kfps | Y |
ae20(2) 36.6kpps |
ae23(1) 33.1kpps |
et-0/0/50 33.1kpps |
N/A |
L2CKT: ACX7100-48L (AN3 to SAG) | 2 | 100kfps | Y | 49.2kpps |
et-0/0/50 53.2kpps |
N/A | N/A |
VPLS: ACX7100-48L (AN3 to SAG) [2] | 3 | 100kfps | N |
ae20(2) 36.0kpps |
ae23(1) 33.2kpps |
et-0/0/50 33.7kpps |
N/A |
EVPN: ACX7100-48L (AN3 to DU) | 3 | 100kfps | N |
ae20(2) 36.0kpps |
ae23(1) 33.4kpps |
et-0/0/50 33.0kpps |
N/A |
L3VPN: ACX7100-48L (AN3 to AG1.2) | 3 | 100kfps | N |
ae20(2) 36.9kpps |
ae23(1) 33.7kpps |
et-0/0/50 33.4kpps |
N/A |
L2VPN: ACX7509 (AG1.1 to SAG) | 2 | 100kfps | Y |
ae4(2) 50kpps |
ae5(1) 49.6kpps |
N/A | N/A |
L2CKT: ACX7509 (AG1.1 to SAG) | 2 | 100kfps | Y |
ae4(2) 53.2kpps |
ae5(2) 50.2kpps |
N/A | N/A |
VPLS: ACX7509 (AG1.1 to SAG) [2] | 2 | 100kfps | N |
ae4(2) 53.3kpps |
ae5(2) 50.1kpps |
N/A | N/A |
EVPN: ACX7509 (AG1.1 to AN1) | 4 | 100kfps | N |
ae4(2) 25.0kpps |
ae19(2) 25.0kpps |
ae22(1) 25.0kpps |
ae23(1) 25.0kpps |
L3VPN: ACX7509 (Unit Test)incrementing src/dst ip x3 | 4 | 50kfps | N |
et-1/0/8 13kpps |
et-1/0/10 12kpps |
et-6/0/8 12kpps |
et-6/0/13 13kpps |
[1] For complete ECMP results with all outputs, contact your account representative.
[2] Only Known Unicast is shown. VPLS BUM traffic should not load balance over ECMP. This is an expected behavior.
Convergence Briefing
Network convergences were validated against the following failure events:
- DUT Adjacent link failures
- DUT Indirect link failures
- DUT Node failures
- Daemon failure and configuration add/delete events
Overall convergence results are within expectations for the given network design. In the Fronthaul (CSR to HSR), failure and restoration events were well within 50ms for all VPN traffic types. The latest generation ACX7000 product family demonstrated optimal convergence, improving upon previous generation Broadcom-based platforms. For all segment routing links, TI-LFA with loose mode allows transitioning to link-protection should node-protection become unavailable.
L2-L3 interworking scenarios with longer convergence can be attributed to control plane triggered global repair events, which may be improved with additional mechanisms such as BGP Prefix Independent Convergence (BGP-PIC). These models include L3VPN to EVPN-ELAN IRB Anycast Gateway, and L3VPN to Bridge Domain with IRB Static Virtual-MAC. EVPN convergence may be further improved with dynamic-list-next-hop (DLNH) and evpn-egress-link-protection (EPL).
Table 2 and Table 3 show the convergence results in the given scenarios:
- TI-LFA without ECMP (default for the ACX7000 Fronthaul segment).
- TI-LFA with ECMP (default for MX/PTX paths in core).
ACX7000 does not simultaneously support FRR and ECMP. During certain path failure events, TI-LFA is prioritized over ECMP to avoid triggering global repair.
Flow Type | 5G EVPN-VPWS (msec) | 5G EVPN-FXC (msec) | 5G L3VPN-BD IRB (msec) | 5G L3VPN-EVPN IRB (msec) | L3VPN (msec) | ||||||
---|---|---|---|---|---|---|---|---|---|---|---|
Events | CSR->DU | DU->CSR | CSR->DU | DU->CSR | DU->SAG | SAG->DU | DU->SAG | SAG->DU | CSR<>HSR | ||
AN3 to AG1.2 disable | 7 | 8 | 0 | 6 | - | - | - | - | 6 | ||
AN3 to AG1.2 enable | 1 | 0 | 0 | 0 | - | - | - | - | 0 | ||
AN3 to AG1.1 disable | 5 | 1 | 0 | 0 | - | - | - | - | 4 | ||
AN3 to AG1.1 enable | 10 | 0 | 0 | 0 | - | - | - | - | 1 | ||
AG1.2 to AG1.1 disable | 5 | 0 | 0 | 0 | 0 | 0 | 5 | 0 | 1 | ||
AG1.2 to AG1.1 enable | 5 | 0 | 0 | 1 | 0 | 0 | 0 | 0 | 1 | ||
AG1.2 to AG2.2 disable | - | - | - | - | 0 | 0 | 0 | 0 | - | ||
AG1.2 to AG2.2 enable | - | - | - | - | 1 | 0 | 0 | 10 | - | ||
AG2.2 to AG3.2 disable | - | - | - | - | 0 | 0 | 0 | 0 | - | ||
AG2.2 to AG3.2 enable | - | - | - | - | 0 | 0 | 1 | 0 | - | ||
CR2 to SAG disable | - | - | - | - | 0 | 0 | 0 | 1 | - | ||
CR2 to SAG enable | - | - | - | - | 1 | 0 | 0 | 0 | - |
Flow Type | 5G EVPN-VPWS (msec) | 5G EVPN-FXC (msec) | 5G L3VPN-BD IRB (msec) | 5G L3VPN-EVPN IRB (msec) | ||||
---|---|---|---|---|---|---|---|---|
Events | CSR<>SAG | HSR<>SAG | CSR<>SAG | HSR<>SAG | CSR<>SAG | HSR<>SAG | CSR<>SAG | HSR<>SAG |
AN3 to AG1.2 disable | 2 | - | 5 | - | 1 | - | 0 | - |
AN3 to AG1.2 enable | 1 | - | 5 | - | 0 | - | 0 | - |
AN3 to AG1.1 disable | 5 | - | 6 | - | 5 | - | 0 | - |
AN3 to AG1.1 enable | 1 | - | 5 | - | 5 | 0 | 0 | - |
AG1.2 to AG1.1 disable | 2 | 2 | 0 | 1 | 2 | 0 | 0 | 4 |
AG1.2 to AG1.1 enable | 2 | 0 | 0 | 5 | 0 | 0 | 0 | 0 |
AG1.2 to AG2.2 disable | 0 | 144 | 4 | 5 | 0 | 4 | 4 | 4 |
AG1.2 to AG2.2 enable | 0 | 1 | 35 | 2 | 0 | 5 | 0 | 0 |
AG2.2 to AG3.2 disable | 0 | 65 | 0 | 80 | 0 | 65 | 0 | 78 |
AG2.2 to AG3.2 enable | 2 | 0 | 55 | 0 | 0 | 2 | 1 | 0 |
CR2 to SAG disable | 0 | 48 | 0 | 60 | 0 | 50 | 0 | 44 |
CR2 to SAG enable | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |