[Paper Review] How Better is Distributed SDN? An Analytical Approach

Ziyao Zhang, Liang Ma+4 more|arXiv (Cornell University)|Dec 12, 2017

Software-Defined Networks and 5G16 references3 citations

TL;DR

This paper presents the first analytical framework to quantify performance gains in distributed SDN networks based on synchronization levels and network structure. Using a generic weighted network model, it derives theoretical bounds for average path length (APL) across four synchronization scenarios—minimum, single, partial, and complete synchronization—demonstrating up to 60% APL reduction with full synchronization and diminishing returns beyond moderate synchronization levels.

ABSTRACT

Distributed software-defined networks (SDN), consisting of multiple inter-connected network domains, each managed by one SDN controller, is an emerging networking architecture that offers balanced centralized control and distributed operations. Under such networking paradigm, most existing works focus on designing sophisticated controller-synchronization strategies to improve joint controller-decision-making for inter-domain routing. However, there is still a lack of fundamental understanding of how the performance of distributed SDN is related to network attributes, thus impossible to justify the necessity of complicated strategies. In this regard, we analyze and quantify the performance enhancement of distributed SDN architectures, influenced by intra-/inter-domain synchronization levels and network structural properties. Based on a generic weighted network model, we establish analytical methods for performance estimation under four synchronization scenarios with increasing synchronization cost. Moreover, two of these synchronization scenarios correspond to extreme cases, i.e., minimum/maximum synchronization, which are, therefore, capable of bounding the performance of distributed SDN with any given synchronization levels. Our theoretical results reveal how network performance is related to synchronization levels and inter-domain connections, the accuracy of which are confirmed by simulations based on both real and synthetic networks. To the best of our knowledge, this is the first work quantifying the performance of distributed SDN analytically, which provides fundamental guidance for future SDN protocol designs and performance estimation.

Motivation & Objective

• To address the lack of fundamental understanding of how network attributes affect distributed SDN performance.
• To quantify the performance improvement of distributed SDN in relation to inter-controller synchronization levels and network topology.
• To establish theoretical bounds for average path length (APL) under varying synchronization scenarios.
• To validate the analytical model through simulations on real and synthetic networks.
• To guide future SDN protocol design by revealing performance tradeoffs between synchronization cost and routing efficiency.

Proposed method

• Proposes a generic weighted network model with m domains, each containing n nodes, and inter-domain gateways modeled as random connections with weight β.
• Derives analytical expressions for APL under four synchronization levels: minimum (MS), single (SS), partial (PS), and complete (CS) synchronization.
• Uses a simplified approximation method based on β^(k-1) to estimate route options in k-hop bus networks, enabling efficient computation of APL for PS and CS.
• Applies probabilistic analysis to estimate the number of available inter-domain paths based on gateway density and synchronization level.
• Validates theoretical predictions via simulations on 30 real and synthetic network realizations with 50 source-destination pairs per realization.
• Uses Corollary 7 to explain convergence of APL to a constant at high β values, indicating diminishing returns in performance gain.

Experimental results

Research questions

• RQ1 How does the average path length (APL) in distributed SDN vary with inter-controller synchronization levels?
• RQ2 What is the theoretical performance bound of distributed SDN under minimum and maximum synchronization?
• RQ3 How do network structural parameters, such as the number of gateways per domain (β), affect routing performance?
• RQ4 To what extent does additional synchronization yield diminishing returns in APL reduction?
• RQ5 Can a theoretical model accurately predict APL in diverse real and synthetic network topologies?

Key findings

• Complete synchronization (CS) achieves the lowest possible APL, serving as a theoretical lower bound, with up to 60% reduction compared to minimum synchronization (MS).
• Single synchronization (SS), which provides only intra-domain link weight information, reduces APL by up to 50% compared to MS, indicating significant benefit from minimal additional information.
• Partial synchronization (PS) shows diminishing returns: increasing τ from 2 to 3 yields minimal APL improvement, especially at low β, indicating a sharp decline in benefit-to-cost ratio.
• As β increases, APL decreases due to higher gateway density enabling more route options, but APL converges to a constant at high β, confirming theoretical convergence as predicted by Corollary 7.
• Theoretical APL estimates closely match simulation results across all synchronization levels and network types, confirming high accuracy of the analytical model.
• Even with minimal synchronization (MS), APL remains significantly smaller than network size (20,000 nodes), highlighting inherent scalability advantages of distributed SDN.