Delay-sensitive In-network Data Aggregation for Server-centric Datacenters
碩士 === 國立臺灣大學 === 資訊工程學研究所 === 101 === Bandwidth consumption minimization is a popular issue for data center networks. For real-time applications, data aggregation is commonly used. Data can be aggregated in a shuffle phase, and the output size can be reduced as a fraction of the input size. Neverth...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | en_US |
| Published: |
2013
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/80487624192237189582 |
| Summary: | 碩士 === 國立臺灣大學 === 資訊工程學研究所 === 101 === Bandwidth consumption minimization is a popular issue for data center networks. For real-time applications, data aggregation is commonly used. Data can be aggregated in a shuffle phase, and the output size can be reduced as a fraction of the input size. Nevertheless, existing works on data aggregation do not consider deadline constraints in datacenters. In this paper, we utilize data aggregation to reduce not only bandwidth consumption but also the latency of dataflows so that the deadlines of dataflows can be satifsified. We also exploit the property in server-centric datacenters (SDC) and choose suitable paths to achieve low end-to-end delay. According to the simulation result, the SDC aggregation-tree algorithm can provide both lower bandwidth consumption and lower delay than previous approaches.
|
|---|