multiCNI

multiCNI refers to the use of multiple container network interface (CNI) plugins within a cluster or pod to provide more than one network path or interface. In practice, the term often describes architectures that combine CNIs so a single workload can access distinct networks, policies, or services through separate interfaces.

The most widely used realization of multiCNI is Multus CNI, which acts as a meta-plugin to attach additional networks to pods using other CNIs. In this model, a primary CNI handles the default network, while Multus delegates secondary networks to other specified CNIs. Network attachments are described by Network Attachment Definitions (NADs), and pods declare the networks they should join through annotations or, in some orchestrators, through API objects. This allows per-pod networking configurations and supports complex topologies such as virtual networks, overlay networks, or direct connections to specific infrastructure networks.

In Kubernetes, multiCNI setups enable scenarios including multi-network orchestration for virtual network functions (VNFs), integration with software-defined networking (SDN) platforms, and isolation of traffic by network, security domain, or QoS requirements. They also support attaching specialized interfaces for purposes like high-performance networking or external connectivity.

Advantages include greater networking flexibility, improved isolation, and the ability to reuse existing CNIs. Potential challenges encompass configuration complexity, management of NADs, debugging difficulties, potential performance overhead, and security considerations arising from multiple network planes. As an architectural pattern, multiCNI is most commonly associated with Multus CNI, but other meta-networking approaches exist in the broader container networking landscape.