Mantis

Abstract

For modern data center switches, the ability to-with minimum latency and maximum flexibility-react to current network conditions is important for managing increasingly dynamic networks. The traditional approach to implementing this type of behavior is through a control plane that is orders of magnitude slower than the speed at which typical data center congestion events occur. More recent alternatives like programmable switches can remember statistics about passing traffic and adjust behavior accordingly, but unfortunately, their capabilities severely limit what can be done.In this paper, we present Mantis, a framework for implementing fine-grained reactive behavior on today's programmable switches with the help of a specialized reactive control plane architecture. Mantis is, thus, a combination of language for specifying dynamic components of packet processing and an optimized, general, and safe control loop for implementing them. Mantis provides a simpleto-reason-about set of abstractions for users, and the Mantis control plane can react to changes in the network in 10s of µs.

Marina: Realizing ML-Driven Real-Time Network Traffic Monitoring at Terabit Scale

“…Our first approach of using the Apache Thrift interface provided by the Tofino SDE failed, as it needed on average 3 s for a complete read of all registers. Thus, we implemented a custom controller library with access to internal functions based on the suggestions of Yu et al [55]. This allows us to circumvent the Thrift layer and achieve a much smaller read duration of 268 ms, independently of the number of monitored flows.…”