ParaFlow: Fine-grained parallel SDN controller for large-scale networks

Using Software-Defined Networking (SDN), the flexibility and programmability of networks can be significantly increased through the decoupling of the control and data planes. However, network scale-up in large-scale data centers can rapidly increase the computational complexity of operations such as...

Full description

Saved in:
Bibliographic Details
Published in:Journal of network and computer applications Vol. 87; pp. 46 - 59
Main Authors: Song, Ping, Liu, Yi, Liu, Chi, Qian, Depei
Format: Journal Article
Language:English
Published: Elsevier Ltd 01.06.2017
Subjects:
C.2.1 [Computer-Communication Networks]: Network Architecture and Design—Network Communications
D.1.3 [Concurrent Programming]: Parallel Programming
Event-based programming
Fine-grained parallelism
Languages
Multithreaded controller
Software-defined networking
Fine-grained parallelism
Software-defined networking
Multithreaded controller
D.1.3 [Concurrent Programming]: Parallel Programming
Languages
Performance
C.2.1 [Computer-Communication Networks]: Network Architecture and Design—Network Communications
Event-based programming
ISSN:1084-8045, 1095-8592
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Using Software-Defined Networking (SDN), the flexibility and programmability of networks can be significantly increased through the decoupling of the control and data planes. However, network scale-up in large-scale data centers can rapidly increase the computational complexity of operations such as the shortest path calculation on the network topology or Quality-of-Service (QoS) routing, which, in turn, can cause scalability problems in current SDN controllers. This paper proposes ParaFlow, a multithreaded SDN controller that supports fine-grained parallelism by exploiting application parallelism and utilizing multi-/many-core resources to accelerate event processing. ParaFlow also provides a flow-based programming interface that allows application developers to program with network flows rather than various types of low-level events. Experimental results show that ParaFlow achieves satisfactory performance and scalability in the multithreaded case.
ISSN:1084-8045
1095-8592
DOI:10.1016/j.jnca.2017.03.009