According to recent reports, the combination of Netflix and YouTube data traffic consist 50% of all Internet bandwidth usage. In an unregulated network, high-data rate applications dominate the competition for bandwidth and this fact may have serious implications for both end-users and telecom operators. More specifically, bandwidth-greedy applications can degrade significantly the user experience for the rest services, or even worse, completely displace low-data rate, real-time applications (e.g., Voice over IP (VoIP) telephony, online games, etc.). From the telecom provider point of view, bandwidth-heavy services create a strong demand for network upgrades, which results to large CAPEX increase.
A fine-grained bandwidth management scheme that offers different types of Service Level Agreement (SLA) guarantees and data prioritization, without noticeable high-/low- data service degradation, is of high value. MENTA aims to validate through experiments an efficient, flexible, and virtualized network bandwidth management solution. MENTA addresses the challenge to provide network visibility (i.e., overview of the data traffic that traverses the network links) through deploying virtualized Deep Packet Inspection (DPI) functions to Fed4FIRE servers. Being aware of the network traffic, MENTA will select the network links that support Quality of Service (QoS) through the usage of Open-Flow enabled switches.
The MENTA experiment-based approach will be composed of two major stages. The first stage will be devoted to validate the traffic classification based on the DPI Virtual Network Function (VNF) within the i2CAT OFELIA island, composed mainly of OpenFlow switches and standard high-volume servers. Once the first stage finishes, the second stage will include the performance evaluation of the QoS-driven bandwidth management scheme, by means of using resources from both the i2CAT Fed4FIRE facility and other interconnected Fed4FIRE islands (i.e., University of Bristol) or emulated machines (i.e., VirtualWall testbed provided by iMinds).