Network convergence in SDN versus OSPF networks
Network availability and reliability remain of primary concern, especially that network failures and changes cannot be completely eliminated, but their consequences can be mitigated and controlled. Accordingly, fast network convergence becomes a significant requirement to control and limit the negat...
Saved in:
Published in | 2018 Fifth International Conference on Software Defined Systems (SDS) pp. 130 - 137 |
---|---|
Main Authors | , , , |
Format | Conference Proceeding |
Language | English |
Published |
IEEE
01.04.2018
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | Network availability and reliability remain of primary concern, especially that network failures and changes cannot be completely eliminated, but their consequences can be mitigated and controlled. Accordingly, fast network convergence becomes a significant requirement to control and limit the negative effects of network failures or changes. Network convergence has been studied for different distributed routing protocols; but with the emergence of centralized routing in SDN, better convergence characteristics are expected. In this paper, we model convergence delays for centralized (SDN) and distributed routing (OSPF), and compare their theoretical behavior to the experimental one. Analytical results show that SDN and OSPF outperform each other under different network conditions, depending on network size, link delays, and fault location. Experimental results verify the theoretical expectations, and provide more insights into convergence delays. Even with minor discrepancies due to controller and switch modeling, the theoretical study and the empirical results give comparable outcomes. |
---|---|
DOI: | 10.1109/SDS.2018.8370434 |