A comparative study of dynamic bandwidth allocation algorithms for long reach passive optical networks
Autor: | Ganesh C. Sankaran, Krishna M. Sivalingam, Anusha Sivakumar |
---|---|
Rok vydání: | 2012 |
Předmět: |
Performance comparison
Dynamic bandwidth allocation Packet networks Computer science Frequency allocation Round-trip time Discrete event simulation Thread (computing) Round-trip delay time Passive optical network DSL Long reach Dynamic bandwidth allocation algorithms CPU resources Electrical and Electronic Engineering Support networks Passive optical networks Packet delay Access network Network packet business.industry Upstream channel ComputerSystemsOrganization_COMPUTER-COMMUNICATIONNETWORKS Discrete events Comparative studies DBA algorithms Digital subscriber line Modems Gap filling Polling business Algorithm Multi-thread Algorithms Computer network |
Zdroj: | IETE Technical Review. 29:405 |
ISSN: | 0256-4602 |
DOI: | 10.4103/0256-4602.103174 |
Popis: | Passive Optical Networks (PON) have emerged as an important fiber-based access network alternative to DSL, Cable, and other access networks. Some of the limitations of PONs are that it can support only up to 64 subscribers and reach up to 20 km. Dynamic Bandwidth Allocation (DBA) algorithms enable multiple subscriber nodes to efficiently share the upstream channel. In this paper, we investigate the performance of DBA algorithms for Long Reach passive optical networks (LR-PON) that support network coverage up to 200 km and up to 2 048 subscribers. Thus, LR-PONs can act as integrated metropolitan and access networks. Due to the increase in round trip time of the network from a few hundred ms in case of PON to few milliseconds in case of LR-PON, DBA algorithms proposed for PONs may not be applicable as is to LR-PONs. The algorithms studied are WDM-IPACT, Intelligent Gap Filling strategy, Multi- thread Polling, and Newly Arrived Frames Plus. The algorithms have been compared in terms of packet delay and consumed CPU resources, using discrete-event simulation modeling. The results indicate that Multi-Thread Polling is most efficient in terms of packet delay however, WDM-IPACT leads to 20% increased delay while consuming only 50% of the CPU resources required for Multi-thread Polling. |
Databáze: | OpenAIRE |
Externí odkaz: |