Virtual networks
have memory.
Since 2019, a research line has been showing that TCP traffic on virtualized networks behaves like a fractal: every combination of operating system, hypervisor and container has its own "signature" of stability and performance over time — and that signature can be measured, compared and used to predict which configuration will keep delivering the best service.
From the dissertation to FRANCISCO 2.0
FRANCISCO began as a theoretical PhD framework and became an open-source web tool. Along the way, three mathematical models were created — WMDEA, SMDEA and Super-Cobb-Douglas — all adaptations of Data Envelopment Analysis (DEA) able to correctly handle ratio-form variables, such as the Hurst parameter and fractal dimension.
Six milestones, one throughline
From a traffic measurement to a recommendation
Measurement
TCP traffic is generated and measured (iperf/iperf3) for each network configuration, treated as a DMU.
Fractal analysis
The fractal dimension and Hurst parameter of each time series are computed.
DEA model
The fractal variables feed a multiplicative DEA model (WMDEA, SMDEA or Super-Cobb-Douglas).
Ranking
Configurations are ranked from most to least efficient, with efficiency frontiers.