Fine-Grained, Multi-Domain Network Resource Abstraction as a Fundamental Primitive to Enable High-Performance, Collaborative Data Sciences

Optimizing in the Dark: Learning an Optimal Solution through a Simple Request Interface

“…The constraints include both physical network constraints (e.g., if two flows share a same link, their allocated bandwidths cannot exceed the link's available bandwidth), and network policies (e.g., rate limiting, etc.) The constraints are captured as an abstraction of linear inequalities (Gao et al 2016;Xiang et al 2018). For example, to capture the physical network constraints, the NNR first retrieves the routes (i.e., sequence of traversed links) for each flow.…”

“…To address this problem, researchers have proposed several solutions, but all of them suffer limitations, and violate privacy requirements. For example, to determine optimal bandwidth reservations, recent proposals depart from the simple reservation interface, and require either networks to reveal sensitive information to users (Soulé et al 2014;Subramanian, D'Antoni, and Akella ;Heorhiadi, Reiter, and Sekar 2016;, or vice versa (Gao et al 2016;Gao et al 2017;Xiang et al 2018). These solutions are therefore limited to settings where the level of trust between the applications and the underlying network is high.…”

Prototipação de Solução de Roteamento de Fonte para Suporte à Engenharia de Tráfego em Ciência de Dados Intensiva

“…Neste contexto, existem várias redes de alto desempenho que interconectam centros de pesquisa, como a ESnet nos Estados Unidos [Xiang et al 2018] e a G ÉANT [Valera-Muros et al 2019] na Europa. Para atender os níveis de QoS (Qualidade de Servic ¸o, do inglês Quality of Service) exigidos para essas redes, devem ser escolhidas ferramentas e métodos que permitam um uso mais eficiente dos recursos disponíveis.…”