Date of Award

Fall 2012

Project Type


Program or Major

Computer Science

Degree Name

Doctor of Philosophy

First Advisor

Elizabeth Varki


Transmitting data via the Internet is a routine and common task for users today. The amount of data being transmitted by the average user has dramatically increased over the past few years. Transferring a gigabyte of data in an entire day was normal, however users are now transmitting multiple gigabytes in a single hour. With the influx of big data and massive scientific data sets that are measured in tens of petabytes, a user has the propensity to transfer even larger amounts of data. When transferring data sets of this magnitude on public or shared networks, the performance of all workloads in the system will be impacted.

This dissertation addresses the issues and challenges inherent with transferring big data over shared networks. A survey of current transfer techniques is provided and these techniques are evaluated in simulated, experimental and live environments. The main contribution of this dissertation is the development of a new, "nice" model for big data transfers, which is based on a store-and-forward methodology instead of an end-to-end approach. This nice model ensures that big data transfers only occur when there is idle bandwidth that can be repurposed for these large transfers. The nice model improves overall performance and significantly reduces the transmission time for big data transfers. The model allows for efficient transfers regardless of time zone differences or variations in bandwidth between sender and receiver. Nice is the first model that addresses the challenges of transferring big data across the globe.