The case for model-driven interpretability of delay-based congestion control protocols

Abstract

Analyzing and interpreting the exact behavior of new delay-based congestion control protocols with complex non-linear control loops is exceptionally difficult in highly variable networks such as cellular networks. This paper proposes a Model-Driven Interpretability (MDI) congestion control framework, which derives a model version of a delay-based protocol by simplifying a congestion control protocol's response into a guided random walk over a two-dimensional Markov model. We demonstrate the case for the MDI framework by using MDI to analyze and interpret the behavior of two delay-based protocols over cellular channels: Verus and Copa. Our results show a successful approximation of throughput and delay characteristics of the protocols' model versions across variable network conditions. The learned model of a protocol provides key insights into an algorithm's convergence properties.

[Solution] ALCC: Migrating Congestion Control To The Application Layer In Cellular Networks

“…However, this can be an exhaustive and time-consuming task, which is out of scope for this paper. Thus, to be able to quickly provide a proof-of-concept implementation of one of the new cellular CC protocols into the ALCC Android library, we relied on using the Model-Driven Interpretable (MDI) congestion control [29] approach. MDI allows approximating any congestion control algorithm as a general discrete-time Markov model by a 2-dimensional state space, represented in the form of a state-transition probability matrix for that algorithm.…”

“…The App relies on MDI as the main congestion control logic. We have utilized a transition probability matrix of Verus that was trained over 1000 cellular traces as described in [29]. Figure 4 shows the throughput and delay performances of the Android ALCC app(using Verus MDI) compared to the TCP cubic performance measured using an upload of the same file with scp 3 from a laptop that was tethered to the Android phone.…”