Overview

Key Features

• Create a network digital twin that is faithful to the physical network at any time: NDTwin updates its perceived network topology whenever a switch or link changes its status. Besides, it uses the sFlow standard, which is supported by all switches, to quickly detect flows emerging or existing in the network and accurately estimate their current bandwidth usages in real time.
• Provide ultra-precision and real-time detection and bandwidth usage estimation of flows: NDTwin uses a very novel technique, which will be published at IEEE International Conference on Communications (ICC) 2026, to achieve ultra-precision sFlow flow sample generation on all links in a hardware network. Currently, due to the control-plane overhead of sFlow on a switch, all switches in the market can only sample a packet out of every 1000, 10000, and 100000 packets (note that these numbers are called the sampling rate in the sFlow standard) on a link when the link speed is 1 Gbps, 10 Gbps, and 100 Gbps, respectively. By using our novel technique, no matter how high the link speed is, one packet can be sampled out of every 100, 10, or even 1 packets arriving on a link. Due to this ultra-precision sampling capability, much more mice flows can be detected in the network and the bandwidth usages of flows can be much more accurately estimated. Note that this capability requires using a P4 hardware switch and this capability is not required for using NDTwin. It is a performance and accuracy booster rather than a requirement. If Mininet is used to form the emulated data network, the sFlow sampling rate for its OVS switches can also be lowered but at the cost of high control-plane overhead.
• Use a digital-twin-powered simulation platform to quickly find the optimal solution to the current or forthcoming problem: NDTwin provides a general-purpose simulation platform where different-purpose simulator programs can be installed and invoked on-demand to simulate many what-if conditions in parallel. Based on the digital-twin-powered simulation results, NDTwin applications can quickly find the optimal solution to the current problem or a forthcoming problem that they are designed to handle.
• Enable an NDTwin application to easily use a trained AI/ML model for inference and optimal control of the network: NDTwin applications can use the network state recorder (NSR) tool of NDTwin to periodically collect and store network data into files for AI/ML model trainning. The trainning can be performed on any machine and based on any AI/ML platform. Like the normal usage of AI/ML model trainning and inference in the real life, the trained model can be exported to a file and then imported into an NDTwin application for real-time inference and optimal control of the network.
• Employ a Software-Defined Network (SDN) controller to real-time control a network based on the digital-twin-found optimal solution: NDTwin can operate a network composed of switches that support the OpenFlow protocol. It uses an SDN controller (e.g., Ryu) to real-time control the network based on the optimal solution found by its digital-twin-powered applications.
• Provide useful digital-twin-powered applications that optimally control a network: NDTwin currently provides serveral useful applications that optimally control a network. The energy-saving application will automatically power off under-utilized switches when the network load drops below a low watermark threshold, thus greatly saving the energy consumption of a network without lowering its Quality of Service (QoS). The traffic-engineering application will dynamically adjust each Equal-Cost Multi-Path (ECMP) group’s flow dispatching mapping, thus balancing the loads of the links of each ECMP group and significantly increasing the overall utilization of a network. More useful digital-twin-powered applications are being developed.
• Provide an open source development platform for any one to develop innovative digital-twin-powered applicatons: NDTwin is open source. Its kernel provides a useful and rich set of API functions for any one to develop innovative digital-twin-powered applications.
• Provide a Web GUI tool (WGT) and real-time network traffic visualizer (NTV): NDTwin provides an easy-to-use Web GUI tool by which the user no longer needs to use Command Line Interface (CLI) to issue network commands, read network settings, or debug network problems. NDTwin also provides a real-time network traffic visualizer (NTV) by which the user can real-time “see” where current flows exist in the network, what routing paths they are taking, and their relative current bandwidth usages.
• Work correctly for both physical networks and emulated networks: NDTwin has been rigorously tested and verified that it works correctly for physical networks composed of tens of hardware switches (Brocade and HPE) and hundreds of hosts using fat-tree-like network topologies. Besides, NDTwin has been rigorously tested and verified that it works correctly for Mininet-emulated networks.
• Provide a large language model (LLM)-powered intent-based network management interface: NDTwin employs LLMs to support intent-based network management. From its Web GUI tool, NDTwin users can use natural language to easily ask NDTwin to retrieve real-time information about a network or request NDTwin to carry out their desired operations.
• Provide a network traffic generation (NTG) framework by which a very large number of flows can be automatically generated and launched without any human efforts: NDTwin provides a powerful network traffic generation framework by which hundreds or thousands of UDP or TCP flows can be automatically generated and launched over time without human efforts among the hosts of a network (which can be either a physical network or a Mininet-emulated network). The characteristics of generated network traffic can be easily set and varied according to a configuration file in which flow size, flow duration, flow protocol, flow sending rate, and other important parameters can be set. This NTG framework is very useful for testing the functions and evaluating the real-life performance of an NDTwin application under development.
• Provide a network state recorder (NSR) that records all network states during a specified interval: NDTwin provides a network state recorder, which can be enabled or disabled as desired to record the states of all network switches and flows during that interval and save the recorded states into files. The Web GUI tool and network traffic visualizer can open these NSR files to replay the activities of the network occurring during the recording interval. This capability is very useful for checking, debugging, and analyzing the behavior of the network and the function of a new NDTwin application under development.
• Is being actively developed and maintained: NDTwin is being actively developed and maintained. More useful digital-twin-powered applications will be added in the future. Besides, its functions, performance, and quality will be continuously enhanced.