Understanding Multi-Path Routing Algorithms in Datacenter Networks
Datacenter is an irreplaceable and crucial infrastructure to power the ever-growing Internet services and applications. In response to today's application constraints (e.g., throughput, end-to-end delay, bandwidth), most datacenter networks are designed to maintain multiple parallel paths betwe...
        Saved in:
      
    
          | Published in | IEEE Global Communications Conference (Online) pp. 1 - 6 | 
|---|---|
| Main Authors | , , , , , | 
| Format | Conference Proceeding | 
| Language | English | 
| Published | 
            IEEE
    
        01.12.2019
     | 
| Subjects | |
| Online Access | Get full text | 
| ISSN | 2576-6813 | 
| DOI | 10.1109/GLOBECOM38437.2019.9013819 | 
Cover
| Summary: | Datacenter is an irreplaceable and crucial infrastructure to power the ever-growing Internet services and applications. In response to today's application constraints (e.g., throughput, end-to-end delay, bandwidth), most datacenter networks are designed to maintain multiple parallel paths between any given pair of hosts. Consequently, multi-path routing has emerged as a technology of choice which can fully utilize the dormant path diversity. However, due to the growing heterogeneity of datacenter topologies and resource requirements, it demands intensive efforts to configure the right multi-path routing algorithms in real deployment according to the specific service targets. Such user burdens are caused by the lack of empirical knowledge about characteristics of various routing algorithms. To fill this gap, we develop a customized simulator DRE based on OMNET++ simulation environment and INET framework, and measure 5 state-of-the-art multi-path routing algorithms in datacenter covering various topologies and metrics. Apart from evaluating the standard macro metrics, we propose three new micro metrics to explore path-level characteristics, which have not been studied before. Our simulator provides a user-friendly interface for users who are interested in datacenter measurements, and the measurement results can promote future multi- path routing algorithm designs. | 
|---|---|
| ISSN: | 2576-6813 | 
| DOI: | 10.1109/GLOBECOM38437.2019.9013819 |