Virtual LAN (VLAN) and Inter-VLAN Routing Performance Analysis
TL;DR
A performance analysis of SVI-based interVLAN routing consistently outperforms Router-on-a-Stick in throughput and latency because routing decisions are made in application-specific hardware rather than in a shared-bandwidth software path, while Router-on-a-Stick retains advantages in initial cost and configuration simplicity for small networks with few VLANs.
Abstract
Virtual Local Area Networks (VLANs) are a foundational mechanism for logically segmenting switched networks into broadcast domains that align with organisational, security, or functional boundaries rather than physical wiring. Because hosts in different VLANs cannot communicate without a Layer 3 device, inter-VLAN routing is required whenever segmented networks must exchange traffic, and network designers commonly choose between two established approaches: Router-on-aStick (RoAS), in which a single router interface carries tagged traffic for multiple VLANs over sub-interfaces, and Switched Virtual Interface (SVI) routing on a Layer 3 switch, in which routing is performed in hardware using an internal virtual interface per VLAN. This paper presents a performance analysis of these two inter-VLAN routing architectures, evaluated through a simulation-based experimental design using packet-per-second throughput, end-to-end latency, jitter, and packet loss as the primary metrics under varying traffic loads. Building on configurations and metrics reported in existing networking literature and vendor documentation, this study develops a structured test topology, a repeatable measurement methodology, and a comparative analysis framework suited to a laboratory or campus-network setting. The analysis indicates that SVI-based interVLAN routing consistently outperforms Router-on-a-Stick in throughput and latency because routing decisions are made in application-specific hardware rather than in a shared-bandwidth software path, while Router-on-a-Stick retains advantages in initial cost and configuration simplicity for small networks with few VLANs. The paper concludes with practical selection guidelines for network administrators and polytechnic-level networking laboratories, and identifies directions for extending the analysis to physical hardware testbeds and higher VLAN counts.