Abstract: Systems, apparatuses and methods provide technology for model generation with intermediate stage caching and re-use, including generating, via a model pipeline, a multi-level set of intermediate stages for a model, caching each of the set of intermediate stages, and responsive to a change in the model pipeline, regenerating an executable for the model using a first one of the cached intermediate stages to bypass regeneration of at least one of the intermediate stages. The multi-level set of intermediate stages can correspond to a hierarchy of processing stages in the model pipeline, where using the first one of the cached intermediate stages results in bypassing regeneration of a corresponding intermediate stage and of all intermediate stages preceding the corresponding intermediate stage in the hierarchy. Further, regenerating an executable for the model can include regenerating one or more intermediate stages following the corresponding intermediate stage in the hierarchy.

Abstract: In general, techniques are described for performing packet loss measurement in a distributed data plane. In one example, a local router includes a plurality of forwarding units that implement a distributed data plane. First and second forwarding units may switch layer two (L2) packet data units (PDUs) between the local router and a remote router using a virtual path. The first and second forwarding may unit may increment, in response to processing any PDU of the PDUs for the virtual path, respective counters stored by the first and second forwarding units. The first and second forwarding units may update, based on the respective counters, a loss-measurement packet (LMP). For instance, the first forwarding unit, upon updating the LMP, may internally forward the LMP to the second forwarding unit. The second forwarding unit, upon updating the LMP, may send the LMP to the remote router.

Abstract: In general, techniques are described for scheduling traffic for delivery over an aggregated bundle of links. A network device comprising an interface and a data plane may implement the techniques. The interface receives packets associated with packet flows. The data plane associates each of the packet flows with a different link of an aggregated bundle of links. The data plane monitors transmission of the packets via the links to determine a representation of an amount of data sent per link. The data plane further determines that bandwidth utilization does not conform to a desired bandwidth utilization based on the determined representation of the amount of data sent per link. The data plane then re-associates the packet flows to different links of the aggregated bundle based on the determination that the bandwidth utilization does not conform to the desired bandwidth utilization.