Hide metadata

dc.contributor.authorKjevik, Eirik Ketilsønn
dc.date.accessioned2019-08-08T23:47:26Z
dc.date.available2019-08-08T23:47:26Z
dc.date.issued2019
dc.identifier.citationKjevik, Eirik Ketilsønn. Active Queue Management for window based applications in Named DataNetworking. Master thesis, University of Oslo, 2019
dc.identifier.urihttp://hdl.handle.net/10852/69056
dc.description.abstractNamed Data Networking (NDN) has built in congestion mechanisms, both in end-node applications and in the forwarding strategy in intermediate nodes. End-nodes may adjust sending rate. Intermediate nodes will rank each face for each Interest Name and store the rating in the Forwarding Information Base (FIB). There is, however a challenge in delegating between different Interest Names on the same face. Therefore, additional Active Queue Management (AQM) methods are required. This thesis investigates a solution to solve the the congested state of the window based application Practical Congestion Control (PCON) in a shared environment with non-adjusting applications by implementation of an AQM in the forwarding strategy. In essence, the thesis show the cross-traffic congestion can be omitted by ensuring the correct flow of Interest packets in both directions. The same direction scenario shows that the intermediate node is able to sort traffic based on the Name attribute and that the PCON are still sensitive to high packet volume. Lastly, it is proven that these two mechanisms can coexist without compromising each other. It shows the limitations of a window-based application when there are non-responsive application in a congested state. A strict intermediate resources delegation, such as the proposed AQM-mechanism, can counteract that one application dominates the network traffic. The proposed AQM method is based on existing Random early discard. It distinguishes between Interest- and Data packet types and ensures that the number of Interest packets allowed upstream are equal to the available bandwidth for Data packets coming back downstream. Fairness is measured as equal amounts of data returning to the end-user back downstream. The Interest packet quota is then divided between each Interest Name-group based on the size of their respective returning Data packet sizes to ensure equal utilisation of link. The performance is analysed in a wide setting. Divided into three different scenarios to focus on the individual components and their properties. To negate the effect of the other NDN congestion avoidance techniques all scenarios were designed to be forced to use a shared bottleneck, this removes the intermediate node forwarding strategy ranked face option. Leaving only the applications to respond to either returning Data packets or to the the NACK packets to change its Interest packet sending rate. The objective was to allow all applications to gain a fair information rate in a congested scenario. The solution is implemented locally in the intermediate nodes on the forwarding strategy, giving high level of control over the elements and does not alter with the underlying structure of ndnSIM. It is limited by operating in intervals meaning it cannot have the buffering effect an intermediate queue would provide.eng
dc.language.isoeng
dc.subjectNamed Data Networking NDN Practical Congestion Control PCON Congestion Control Active Queue Manangement AQM Content Centric Networking CCN
dc.titleActive Queue Management for window based applications in Named DataNetworkingeng
dc.typeMaster thesis
dc.date.updated2019-08-09T23:45:54Z
dc.creator.authorKjevik, Eirik Ketilsønn
dc.identifier.urnURN:NBN:no-72197
dc.type.documentMasteroppgave
dc.identifier.fulltextFulltext https://www.duo.uio.no/bitstream/handle/10852/69056/1/MasterThesis_v5_1.pdf


Files in this item

Appears in the following Collection

Hide metadata