Engineering and evaluating bulk data transfer planning over wide area networks

Pandora;Bulk Data Transfer Planning;Linear Programing;graphic processor unit (GPU);Web Service;Representational\rState Transfer (REST);C++ Framework;Shipping Networks;Cloud Computing

Krueger, Simon P. ; Gupta ; Indranil

关键词: Pandora;    Bulk Data Transfer Planning;    Linear Programing;    graphic processor unit (GPU);    Web Service;    Representational\rState Transfer (REST);    C++ Framework;    Shipping Networks;    Cloud Computing;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/24184/Krueger_Simon.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Users of cloud computing services are faced with the problem of planning and transferring large amounts of data across a geographically dispersed network under time and budget constraints. Pandora is the first service to solve the planning problem but underneath it has a set of issues, which are addressed in this work. This thesis enhances Pandora by improving its interface, scalability, modularity, and extensibility. This is done by evaluating the performance of various GPU linear program solvers on Pandora’s workloads, and by designing and implementing two new systems: a modular C++ framework called Pandora’s Toolbox that solves the bulk data transfer problem, and a web service for planning bulk data transfers. Pandora’s Toolbox models the steps of the bulk data transfer problem and has an extensible design that allows various linear program solvers (e.g., GLPK and GPU solvers) to integrate into it. The web service effectively allows users to construct a shipping network and plan bulk data transfers over it through both a web and a RESTful interface.