Caching over-the-top services in an ISP's network
Translated title
At cache over-the-top tjenester i en ISPs netværk
Author
Jensen, Stefan Almind
Term
4. term
Education
Publication year
2014
Pages
101
Abstract
Video og andet over-the-top (OTT)-indhold leveret via det åbne internet udgør cirka 67% af downstream-trafikken på fast adgang i spidsbelastningsperioder i Nordamerika. For at håndtere denne efterspørgsel placerer OTT-udbydere caches—servere der lagrer populært indhold tæt på brugerne—inde i internetudbyderes (ISP) net. Denne afhandling foreslår en praktisk metode til at planlægge, hvor sådanne caches skal placeres i et ISP-netværk, så backhaul-forbindelserne aflastes. Målet er at reducere trængsel ved at minimere den maksimale belastning på et hvilket som helst link, hvilket hjælper med at balancere trafikken på tværs af netværket. Metoden anvender en genetisk algoritme (en optimeringsteknik inspireret af naturlig selektion) kombineret med en routingsmodel og operatorer, der er målrettet Link Load Balanced Capacitated Facility Location (LLB-CFL)-problemet—at vælge cacheplaceringer med kapacitetsbegrænsninger samtidig med at linkbelastninger balanceres. For at skabe realistiske resultater bruges GIS-data fra Bornholm, Danmark. Efterspørgsel og cachespecifikationer kommer fra et case-studie af Netflix. Evalueringen sammenligner netværket før og efter udrulning af caches, undersøger afvejningen mellem omkostning og fitness (optimeringsydelse), og rapporterer linkbelastninger for trafik, der betjenes af caches og af gatewayen. Resultaterne viser, at den foreslåede metode kan finde cacheplaceringer, der i væsentlig grad aflaster og balancerer netværket.
Video and other over-the-top (OTT) services delivered over the open internet account for about 67% of fixed-access downstream traffic during peak hours in North America. To handle this demand, OTT providers place caches—servers that store popular content close to users—inside internet service providers’ (ISP) networks. This thesis proposes a practical way to plan where to place such caches in an ISP network so that backhaul links are relieved. The goal is to reduce congestion by minimizing the maximum load on any link, which helps balance traffic across the network. The method uses a genetic algorithm (an optimization technique inspired by natural selection) combined with a routing model and operators tailored to the Link Load Balanced Capacitated Facility Location (LLB-CFL) problem—choosing cache locations with capacity limits while balancing link loads. To produce realistic results, the study uses Geographic Information System (GIS) data from Bornholm, Denmark. Demand patterns and cache specifications are taken from a Netflix case study. The evaluation compares the network before and after deploying caches, examines the trade-off between cost and fitness (optimization performance), and reports link loads for traffic served by caches and by the gateway. The results show that the proposed approach identifies cache placements that significantly offload and balance the network.
[This abstract was generated with the help of AI]
Keywords
Genetic algorithm ; Cache ; Over-the-top ; OTT ; GA ; Network planning ; network ; ISP ; link load ; balance ; link load balanced ; fitness ; crossover ; mutation ; genetic ; planning ; routing algorithm ; routing ; ring ; double ring ; maximum link load ; minimise ; cdn ; netflix ; youtube ; offload ; gateways ; caches ; caching ; Internet ; optimisation ; combinatorial optimisation ; facility location ; capacitated ; demand ; genes
Documents