UC Santa Cruz

Despite the vast set of prior work on identifier-locator split architectures, no one approach has seen much success, adoption, or deployment in the Internet at large. We identify the key set of challenges that have inhibited the deployment of these proposals to date, and introduce the Dynamic Internet Mobility for End-Systems (DIME) approach. DIME is based on dynamic address translation between the transport and network layers of end hosts, combined with a simple out-of-band protocol that updates host-address bindings as needed. DIME is the first and only proposal that achieves a clean identifier-locator split without  requiring modifications to the end-host OS or applications; modifications to existing network protocols, security mechanisms, or hardware; or  a new host-identifier namespace. We  evaluate a Linux daemon implementation of DIME, and show that it i outperforms existing mobility proposals such as mobile IP (MIPv6), multipath TCP (MPTCP), and the Host Identity Protocol  (HIP) across a wide range of performance metrics.