load-balancing-options (Aggregated Multiservices)
Statement introduced in Junos OS Release 11.4.
Support added in Junos OS Release 19.3R2 for Next Gen Services on MX Series routers MX240, MX480 and MX960 with the MX-SPC3 services card.
Configure the high availability (HA) options for the aggregated multiservices (AMS) interface.
Many-to-one (N:1) high availability mode for service applications like Network Address Translation (NAT) is supported. In the case of N:1 high availability mode, one services PIC is the backup (in hot standby mode) for one or more (N) active services PICs. If one of the active services PICs goes down, then the backup replaces it as the active services PIC. When the failed PIC comes back online, it becomes the new backup. This is called floating backup mode. In an N:1 (stateless) configuration, traffic states and data structures are not synchronized between active PICs and the backup PIC.
You can also configure a one-to-one (1:1) high availability mode. In the 1:1 configuration, a single interface is configured as the backup for another single active interface. If the active interface goes down, the backup interface replaces it as the active interface. A 1:1 (stateful) configuration synchronizes traffic states and data structures between the active services PIC and the backup services PIC. This is required for IPsec connections. One-to-one high availability is supported on the MS-MPC but it is not supported for MX-SPC3 in this release.
Load-balancing might not be uniform among member interfaces in certain network deployments. The variance can be because of a misconfiguration, which causes the traffic itself not to be sufficiently randomly distributed, causing the hash keys to be ineffective (for example, the hash key is destination IP but all sessions have only source IP address). The variation can be within the expected range and the load balancing depends on the IP addresses chosen. The hash calculation performs a checksum on several bits of the IP address and not only on the last few lower significant bits of the IP address. In such a scenario, the load-balancing ratio can change, for instance, if the source IP address is changed from 18.104.22.168/24 to 22.214.171.124/24.
The distribution of traffic across member interfaces of an AMS interface is static load-balancing. Flows are load balanced based on a packet hash on parameters such as source IP or destination IP. Load-balancing effectiveness depends on the IP address or protocol diversity. For example, if the hash key is destination IP and all packets have the same destination, then all flows are directed to the same member. This is flow-level load balancing and not per packet. As a result, traffic between a pair of addresses may be 10,000 pps, whereas another pair of addresses may have 1 pps. The load of the former is not distributed among members. High availability is limited to stateless HA. When a backup interface takes over as an active interface, all flows are reestablished (for example. packets may undergo NAT processing differently after failover).
With a stateful firewall, static NAT as basic-nat44 or destination-nat44, and dynamic NAT as nat64, napt-44, dynamic-nat44, and with application layer gateways (ALGs) configured, NAT hairpinning is not supported. Input direction for rule match to be applied is supported only for dynamic NAT types (NAT64, NAT44, and dynamic-NAT44). Service-set policies need to have input or input-output direction only. Flows on all active members are reset when the number of actives changes. The resetting of flows can be avoided at the cost of failed-member's traffic loss using certain options.
The remaining statements are explained separately. See CLI Explorer.
Required Privilege Level
interface—To view this statement in the configuration.
interface-control—To add this statement to the configuration.