ARGON: Reservation in Grid-enabled Networks

DFN-Forum Kommunikationstechnologien(2008)

引用 26|浏览20
暂无评分
摘要
Grid computing offers heterogeneous and distributed resources to scientific communities. Apparently, networks connecting these resources can also be consid- ered as Grid resources. This paper presents ARGON, a system that integrates metro and wide area networks into Grid environments by providing advance reservations and guaranteed network services. Here, single-domain as well as multidomain net- work environments are considered. A major objective is to support metaschedulers in the planning of workflows for e-science applications with demanding network require- ments. According to (Fer07), the analysis of local and remote data becomes more and more im- portant in the field of e-science, medicine, engineering, and digital art. These data sets or streams may be as large as several terabytes, and may be real-time or preprocessed. It is important to guarantee that the data to be processed are delivered timely. It is envisioned that requests for multiple Gbps need to be fulfilled in order to leverage these applications. Since currently this requested quality of service (QoS) cannot be ensured in the Internet, many Grid sites are additionally interconnected by high-speed networks. These networks usually just provide manually arranged connections with a given QoS. However, even in a high-speed network environment it might not be feasible to overpro- vision the network links, and the available bandwidth must be shared and allocated in an efficient manner. Furthermore, a static network setup restricts the selection of available Grid resources. In addition to high bandwidth requirements, interactive and collaborative environments may require specific round trip times, a low packet loss, and small jitter or a combination of those. In order to allow for a flexible assortment, a system is needed that allocates network paths
更多
查看译文
关键词
single domain,round trip time,real time,packet loss,grid computing,quality of service
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要