Traffic can be placed on different transports — such as broadband, cellular, Multiprotocol Label Switching (MPLS), or satellite — to increase usable capacity, improve availability, or satisfy application policies. The precise behavior depends on the product, service, and aggregation endpoints.
Methods range from active/standby failover and per-flow load distribution to packet-level bonding across several paths. Per-flow methods keep each flow on one path, while packet-level methods may spread one flow across paths and then resequence packets. The latter generally requires compatible endpoints or an aggregation service and must handle different latency, loss, maximum transmission unit, and capacity characteristics.
Key points
ObjectiveDefine whether the design is intended for resilience, aggregate throughput, single-flow acceleration, cost control, or application steering.
Path controlMeasure links and specify failure detection, congestion handling, path diversity, and session persistence for each traffic-distribution mode.
EndpointsPacket-level aggregation requires suitable endpoints or a remote service that can resequence traffic and accommodate unequal paths.
Important limitationNominal link speeds are not simply additive. A single flow may remain bounded by one path, packet reordering can reduce performance, and links bought from different providers may still share physical infrastructure. Aggregation also provides neither encryption nor security inspection by itself.