An adaptable workload-agnostic flow scheduling mechanism for Data Center Networks

Cloud applications are an important phenomenon on the modern use of Internet. Search engines, social networks, content delivery and retail and e-commerce sites belong to this group of applications. These applications run on specialized facilities called data centers. An important element in the arch...

Full description

Autores:: Gutiérrez Betancur, Sergio Armando

Tipo de recurso:: Doctoral thesis

Fecha de publicación:: 2018

Institución:: Universidad Nacional de Colombia

Repositorio:: Universidad Nacional de Colombia

Idioma:: spa

Description
Summary:	Cloud applications are an important phenomenon on the modern use of Internet. Search engines, social networks, content delivery and retail and e-commerce sites belong to this group of applications. These applications run on specialized facilities called data centers. An important element in the architecture of data centers is the communication infrastructure, commonly known as data center network (DCN). One of the challenges that DCN have to address is the satisfaction of service requirements of the applications expressed in terms of high responsiveness and high performance. In order to address this challenge, the traffic associated to these applications needs an special handling due to its properties which makes it essentially different to the traffic of other Internet applications such as mail or multimedia services. In order to contribute to the achievement of the previously mentioned performance goals, DCN should be able to prioritize the short flows (a few KB) over the long flows (several MB). However, given the time and space variations that the traffic presents, the information about flow sizes is not available in advance in order to plan the flow scheduling. In this thesis we present an adaptable workload-agnostic flow scheduling mechanism called AWAFS. It is an adaptable approach capable to agnostically adjust the scheduling configuration within DCN switches. This agnostic adjustment contributes to reduce the Flow Completion Time (FCT) of those short flows representing around 85 % of the traffic handled by cloud applications. Our results show that AWAFS can reduce the average FCT of short flows up to 24 % when compared to an agnostic non-adaptable state-of-the-art solution. Indeed, it can provide improvements of up to 60 % for medium flows and 39 % for long flows. Also, AWAFS can improve the FCT for short flows in scenarios with high heterogeneity in the traffic present in the network with a reduction of up to 35 %.

An adaptable workload-agnostic flow scheduling mechanism for Data Center Networks

Publicaciones similares