Bujaranpally, Krishnaveni. Design and implementation of the MobilityFirst protocol stack on an SDN platform using ipv6. Retrieved from https://doi.org/doi:10.7282/T3W66P2P
DescriptionThis thesis presents the design, implementation and evaluation of the MobilityFirst (MF) future internet architecture (FIA) on a Software Defined Network platform using IPv6 header fields. Some of the key features of the MobilityFirst architecture include unique identifiers (GUID) which separate naming from addressing, hop by hop transport protocol, in network storage and computation, global name resolution service (GNRS) which contains mappings of GUIDs to Network Addresses, a storage aware intra do- main routing protocol and an edge aware inter domain routing protocol. The prototype implementation has been done on an SDN platform using the Open source Floodlight controller and Open vSwitches supporting OpenFlow 1.3 protocol. The prototype en- capsulates all the aforementioned features of the MobilityFirst architecture. In order for the OpenFlow protocol to effectively handle MobilityFirst packets, the host stack of the MobilityFirst architecture has been modified to map the source and the destination GUID in the routing header to the source and the destination address fields of the IPv6 header of the packet. The controller then leverages this mapping to identify the source and destination and installs flow rules in the switches to route packets for hosts in the same domain without the help of GNRS. For the source and the destination in different domains, the controller obtains the network address of the destination network from the GNRS and uses the nSPs to identify the appropriate border router to forward packets. Additional modules have been added to the Floodlight controller to analyze and set up flow rules for the MobilityFirst packets and to communicate with the GNRS. In order to implement the in-network storage capability for an SDN an API has been devel- oped for the controller to interact with the storage device. The controller routes the chunks to the storage box when the destination is not reachable or when there is host mobility. The prototype has been evaluated for throughput and latency by conducting experiments on the ORBIT test bed. The processing time of the controller is in the order of few milliseconds and the throughput reaches 750 Mbps and 600 Mbps for 1 MB chunks when the source and the destination are in the same domain and different domains respectively. The latency introduced by the storage box depends on the time interval between the GNRS queries made by the controller and increases as the query interval increases.