MPT-GRE: A Novel Multipath Communication Technology for the Cloud
Szabolcs Szilágyi, Imre Bordán, Lajos Harangi, Benjamin Kiss · 2018
The Cognitive Infocommunications became a hot research area during the past years and it is becoming more and more prevalent. For this reason, telemedicine and the financial services industry are becoming more and more widespread, relying on cloud services. The underlying network infrastructures for the cloud are provided by data centers, which needs to be able to satisfy the new requirements set forth by these systems. The currently deployed data center networks usually have multipath data transfer connections to provide inter-server physical redundancy. Using traditional TCP, they cannot utilize the advantages offered by path aggregation, e.g. improved throughput performance. However, during data center operations both throughput-oriented longer dataflows and delay-sensitive applications can occur. In the case of the latter, even the most representative multipath solution, Multipath TCP (MPTCP), performs poorly, because of its operational architecture. In this paper we present a new L3 multipath communication technology that is capable of efficient data transfer both over TCP and UDP in a dual-path Gigabit Ethernet environment. The MPT network layer multipath communication library - which was proposed to be a possible new basis for the future Cognitive Infocommunications - is capable to use multiple communication channels to create an UDP tunnel which uses GRE tunnel protocol, improving the throughput performance of the cloud-based data centers.