22 - STUDENT - Real Time Quality of Service Pathfinder

Daniel Guija Alcaraz (Polytechnical University of Catalunya)

The work presented in this paper is based on using the Software Defined Networking (SDN) concept to design and develop algorithms and mechanisms to achieve end-to-end Quality of Service (QoS) for real-time applications. The algorithm (Pathfinder) is responsible for the dynamic and on demand provisioning of network resources upon changes in the application traffic requirements. To this target, we make use of the standardized OpenFlow [OpenFlow] protocol which allows the separation of the control and data planes and supports basic ways to offer QoS. The solution builds up an SDN Application (SDNApp) based on the Network as a Service paradigm (Naas) to provide Real Time QoS over an Openflow-based network, adapting the control plane to satisfy the requirements of an application or service.
The PathFinder SDNApp will be embedded in the Network Control Layer (NCL) [NCL] developed within the FP7 OFERTIE project [OFERTIE] which has been implemented based on the OpenNaaS framework [OpenNaaS]. OpenNaaS confers a sophisticated manner to supply applications with on-demand self-service, resource manipulation rights, resource pooling, flexibility, elasticity and dynamic service management and allows the virtualization of the network resources. The Pathfinder SDNApp is an external module designed to work along with an SDN Controller (the NCL) and is based on its specific techniques to provide and maintain end-to-end QoS. It makes use of queuing mechanisms to match certain parameters of a flow to a specific queue on a port. The goals for this proposal include:
- Provide end-to-end on demand QoS over an Openflow network for application and services requests- Search feasible paths meeting QoS requirements and choose the optimal path to assure required Quality of Service.
The PathFinderSDNApp integrated within the OFERTIE project works with the OpenNaaS NCL logic. The SDNApp follows the Flow diagram procedure shown in the poster.
The approach taken in this proposal is to use a Less-greedy algorithm. While other QoS algorithms are purely shortest path based, our algorithm considers the existence of Best Effort traffic. This means that the algorithm not only tries to find an optimal path, but also takes into account network-wide traffic implications using link state and port monitoring mechanisms. Currently, it considers bandwidth and avoids to route through network links that could be easily congested. To achieve this, an Event triggering mechanism is used based on link
utilization/throughput data gathered by the network monitoring. It will detect flow threshold violations to avoid congestions and minimize delay and packet-loss rates. To obtain the preliminary results we set up a network topology (showed in Scenario example) using Mininet tools. Then, the QoS algorithm is applied to User – Multimedia Server requests, obtaining QoS paths (green flow in poster’s Scenario example). Two queue types are created, one to guarantee a minimum bit rate for QoS traffic and the other to limit Best Effort traffic.
UDP bidirectional traffic is transmitted between User – Multimedia Server and UDP bidirectional traffic is transmitted between Other’s users hosts (blue flow in the Scenario example) using Iperf tools to simulate “QoS traffic” and “Best Effort traffic”respectively. The jitter and bandwidth is measured, showing that without configuring queues, maximum and minimum peaks go from 24 msec to 4,5msec for jitter variation for User – Mutimedia Server transfers. While configuring queues, the results show more stable numbers, according to minimum and maximum queue bit rates for User – Multimedia Server transfers.

Download file