Dandelion: A Novel, High-Level Programming System for Software Defined Coalitions with Local State Sharing

Integrating software defined networking into military coalition to realize a software defined coalition (SDC) infrastructure is non-trivial due to the stringent requirements (e.g., real-time, efficiency and reliability) of SDC applications in highly dynamic tactical network environments. Recent designed SDC datapaths improve the performance of SDC applications by offloading complex, stateful operations from the SDC control plane to data plane devices. However, they have two limitations. First, results of offloaded operations (referred to as local state) are not shared between data plane devices, resulting in substantial performance issues in SDC networks. Second, configuring low-level SDC datapaths is time-consuming and error-prone. To address these two issues, in this paper, we design Dandelion, a novel, high-level SDC programming system that allows users to specify the behavior of SDC data plane devices and the datapath configurations can be automatically generated with the local state sharing to efficiently utilize the network resources. We implement a prototype of Dandelion and demonstrate its efficiency and efficacy using experiments. Results show that with Dandelion, the total throughput is two times higher than that of without Dandelion.