Network supported congestion avoidance in software-defined networks

Abstract:

Large IP networks tend to have a highly heterogeneous structure with many different transmission segments, each endowed with individual characteristics like data rate or delay. Obviously, applications find it difficult to precisely adjust their sending behavior to the underlying topology. Basically, a higher network utilization without congestion is desirable. Server applications demand for reliable information about the available data rate in the network in order to provide a sufficient quality to the clients. Traditionally, this information is completely gathered by source and sink without interaction with the network infrastructure. The rise of software-defined networks (SDNs) enables the integration of the network into this process. In this paper we propose a rate control mechanism for software-defined networks to avoid congestion situations. The approach enables clients to communicate with the network to bidirectionally exchange information about available and occupied data rate. We found that this customized communication approach enables better network load distribution and application transmission reaction - which can be easily implemented with SDNs. Ultimately, the controller finds an efficient forwarding path in the network and the clients have a reliable upper bound for their sending data rate.