A mailing list improves communication for all the people interested in a 01.org Project: developers, final users and contributors. A more direct channel to solve your questions and help others to get involved.
The Intel Access Network Dataplanes project will host reference implementations of access network data-planes which are optimized for high-performance packet processing on Intel® architecture, and which are deployable on an NFV platform.
The first of these is a vCMTS (virtualized Cable Modem Termination System) Reference Dataplane which is based heavily on DPDK. This is described in greater detail below.
Intel vCMTS Reference Dataplane and NFV Stack
Intel has developed a DOCSIS MAC dataplane pipeline which is based on DOCSIS 3.1 specifications (MULPI, DEPI, UEPI and SEC) and the DPDK packet-processing framework. The key purpose of this development is to provide a tool for characterization of vCMTS dataplane packet-processing performance and power-consumption on Intel® Xeon® platforms. An overview of the DPDK-based upstream and downstream DOCSIS MAC dataplane developed at Intel is shown below.
The release package provided on this site enables installation of a vCMTS reference dataplane environment consisting of a vCMTS dataplane node and a traffic generation node, both of which are fully containerized and orchestrated by Kubernetes*. An overview of the Intel vCMTS reference dataplane platform is shown below.
On the vCMTS dataplane node, multiple Docker* containers host DPDK-based DOCSIS MAC upstream and downstream dataplane processing for individual cable service-groups, allowing them to be instantiated and scaled independently. On the CMTS traffic-generation node, Docker containers host DPDK Pktgen-based traffic generation instances which simulate upstream and downstream traffic into corresponding vCMTS dataplane instances.
Under Kubernetes, each vCMTS dataplane POD represents a service-group and has separate containers for upstream and downstream DOCSIS MAC dataplane processing. DOCSIS control-plane is simulated through a JSON file containing subscriber cable-modem information. DOCSIS upstream scheduling is simulated through generation of UEPI-encapsulated cable-modem DOCSIS stream segments using PCAP files.
Telemetry functions run in Docker containers as a Daemonset under Kubernetes. A comprehensive set of vCMTS dataplane statistics and platform KPI’s are gathered by the open-source Collectd daemon and a Grafana* dashboard is provided for visualization of these metrics via InfluxDB*. The reference platform also contains a Power-Manager which reduces power-consumption of CPU cores during quiet network periods.
The entire system is orchestrated by Kubernetes, which is an open source container manager and orchestrator. It automates the deployment, scaling, and operational functions associated with application containers. Intel plugins for Kubernetes are used for infrastructure management functions such as CPU core management and assignment of SR-IOV interfaces for NIC’s and QAT devices.
Kubernetes* and Docker* may also be disabled so that the vCMTS dataplane and telemetry processes may run on bare-metal. In this case, process management is performed by the vcmts-pm (platform-management) tool as shown below. Such an environment may be preferable for dataplane performance analysis and benchmarking.