Preprint

Preprint Reports Core Connectivity and Shortest Paths in SDN Test

A virtual model of Unsoed's three-router core reports interconnected switches, attached hosts and shortest-path forwarding, while physical and link-failure tests remain ahead.

A preprint reports that a virtual test of a university network's core layer found all switches interconnected and each host attached to its switch, while test packets followed the shortest or best paths in the layout being examined. The work modeled Universitas Jenderal Soedirman's core network. OSPF calculated the paths in virtual routers, while RFProxy applied the routing information to the Mininet data plane. The study therefore describes the behavior of one specified virtual setup.

The network behind the test

The experiment used two virtual machines. One ran the SDN/OpenFlow data plane, the virtual switch-and-traffic side, with Open vSwitch in Mininet, and the other ran the control plane with RouteFlow and POX. The proposed core used a triangle layout with 1 Gbps links, OSPF routing and RouteFlow with a POX controller. In other words, the setup kept the virtual switches and the routing software on separate machines.

The case study represented Unsoed's core layer, linking three routers at Duren Tiga, Grendeng and Blater with 1 Gbps links. The paper evaluated SDN/OpenFlow and RouteFlow with OSPF in that triangle layout, using interconnectivity and packet-routing tests. Its question was whether the components could connect the modeled hosts and send packets along the paths selected in the model.

Two ways to check the design

The evaluation comprised two validation tests: interconnectivity and packet routing. For the interconnectivity check, the authors used ping on three directed host pairs, h1 to h2, h2 to h3 and h3 to h1, repeating each pair 10 times. The resulting ping series supplied round-trip times, or RTTs, meaning the recorded time for a test to make a return trip. For the routing check, traffic was monitored across all switch interfaces with tcpdump. The captured data were analyzed with Wireshark. The analysis also included OpenFlow packet observations.

The reported average RTT entries were 49.841, 40.644 and 65.129. The authors summarize those averages as a 40 to 65 millisecond range and describe it as normal, while also saying it was higher than local-area network, or LAN, RTT values. The paper presents the numbers as averages for the three listed host-pair series, rather than as one overall network value.

What the packet traces showed

On connectivity, the authors report that all network switches were interconnected and each host was connected to its switch. For route selection, they report that packets for h1-h2, h2-h3 and h3-h1 followed the shortest or best path. OSPF computed those paths in the virtual routers, and RFProxy applied the routing information to the Mininet data plane. The result is a report of forwarding in the tested triangle.

The interface captures offered a second check on that account. Activity appeared on s1 and s2 for h1-h2, on s2 and s3 for h2-h3, and on s1 and s3 for h3-h1. Those interface patterns matched the switch pairs associated with the routes reported for the three host pairs, giving qualitative agreement between the captures and the route description.

The work still ahead

The study's boundary is set by the environment it tested. The work used virtual machines, Mininet, Open vSwitch, RouteFlow and POX, and the case was the specified three-router core. The authors list physical SDN testing as future work, along with tests of availability under link failure, RTT and convergence time, and comparisons with state-of-the-art approaches. Those plans would extend the evaluation beyond the model.

The authors' future-work list makes the scope explicit: physical implementation, failure availability, convergence measurement and comparative testing remain open. Until those checks are made, the reported RTTs and routes belong to this virtual triangle and its three host-pair tests.

The acknowledgements thank LPTSI Unsoed staff for providing details of the university's campus network, particularly its core layer. The supplied metadata classifies the work as a preprint. The paper's conclusion is limited to the tested setup: the modeled switches and hosts were connected, and the modeled traffic followed the OSPF paths reported for it, with broader validation listed as future work.

Paper data and sources

Original title: Designing Core Layer in Campus Network Using Software-Defined Networking
Authors: Iwan Setiawan, Azis Wisnu Widhi Nugraha, Sauqi Asrorul Zaza
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-26
DOI: Not available
Original paper · Full text

Versions and corrections

  1. Published automatically after legal-source, freshness, evidence, and independent-verification gates passed.