In this paper, we conduct the first global-scale measurement study to unveil how 30 mobile operators manage mobility support in their carrier networks. Using a novel, device-centric tool, MMLab, we are able to crawl runtime configurations without the assistance from operators. Using handoff configurations from 32,000+ cells and > 18,700 handoff instances, we uncover how policy-based handoffs work in practice. We further study how the configuration parameters affect the handoff performance and user data access.

Our study exhibits three main points regarding handoff configurations. 1) Operators deploy extremely complex and diverse configurations to control how handoff is performed. 2) The setting of handoff configuration values affect data performance in a rational way. 3) While giving better control granularity over handoff procedures, such diverse configurations also lead to unexpected negative compound effects to performance and efficiency. Moreover, our study of mobility support through a device-side approach gives valuable insights to network operators, mobile users and the research community.