Longitudinal view of ECN adoption
To study long-term ECN adoption, we analyzed passive measurements from both transit networks (MAWI) and client networks (M-Lab)
Longitudinal results on network support for ECN
A Transit-network Perspective (MAWI)
-
Using MAWI traces spanning 2007–2025, we examined the evolution of ECN deployment from a transit-network perspective. We observed a gradual increase in ECN negotiation over time, particularly after 2017, indicating growing deployment across the Internet. While ECN-enabled traffic remained a relatively small fraction of all TCP flows, its prevalence increased steadily throughout the observation period. To determine whether ECN was actively used beyond connection establishment, we further analyzed flows of different sizes and found that ECN was more common in larger, data-carrying flows. This suggests that ECN is increasingly being used by real applications rather than being limited to experimental or measurement traffic. Additional analyses and results are provided in the paper.
Client-networks Perspective (M-Lab)
-
We complemented the transit-network analysis with a client-network perspective using M-Lab NDT measurements collected between 2021 and 2025. By analyzing ECN negotiation and usage signals from residential ISP and mobile networks, we observed a general increase in ECN negotiation over time across both IPv4 and IPv6. However, successful ECN usage often lagged behind negotiation rates, particularly in mobile networks, suggesting that while endpoint support for ECN is becoming more common, support along network paths remains incomplete. We also found that IPv6 networks were consistently more ECN-friendly than their IPv4 counterparts, exhibiting higher negotiation and successful ECN usage rates. Despite the widespread ECN-awareness observed in our active measurements and the relatively low prevalence of ECN bleaching, end-to-end ECN deployment remained limited in many client networks. Additional analyses and breakdowns by network type are provided in the paper.