Both these solutions are built with Angular, GraphQL, Python applications running on Intel / ARM CPU's of the white-box switch control-planes
Telemetry and the capability to pull meaningful data are key for efficient ML driven network analysis. GraphQL paired with Telegraph push and Prometheus pull models gives you the capability to efficiently stream and query metric for network overview and detailed analysis workflows.
Nearly a decade ago I had championed a project to integrate the router/switch forwarding planes with identity. Identity being a user or device/end-point level object using which you could define interaction rules between these objects, starting with simple ones likes ACL and QoS. This would allow for networking and security user policy level abstractions which could be easily modeled and validated. Implementation was an out of the box approach to have multiple data-plane identity source paths contexts wherein existing identity unaware features/services could easily be applied to these different paths.
Over the years this concept has evolved into Intent Based Networking and Zero-Trust network concepts that leverage these same building blocks but there is lots that still needs to be done. With the CrossFlow (simultaneous Traditional + OpenFlow networking) architecture I continue pushing for this integration of networking and security policies.
Currently prototyping k3s orchestrated control-plane service pipelines for the new line of ODM Broadcom Trident III switches with multi-core CPU and 10G Ethernet paths to the ASICs
How is a switch dual control-plane programmability any better than a traditional control-plane - you can filter and duplicate traffic without any impact to regular traffic flow - you don’t reset the entire interface – you don’t disrupt the existing traffic - speed and efficiency in programmability - open standard Open-flow programming constructs - interoperability – applications can be deployed without having to reinvent networking
