Doctor of Philosophy (PhD)


Computer Science

Document Type



Over the years, scientific applications have become more complex and more data intensive. Although through the use of distributed resources the institutions and organizations gain access to the resources needed for their large-scale applications, complex middleware is required to orchestrate the use of these storage and network resources between collaborating parties, and to manage the end-to-end processing of data. We present a new data scheduling paradigm with advance reservation and provisioning. Our methodology provides a basis for provisioning end-to-end high performance data transfers which require integration between system, storage and network resources, and coordination between reservation managers and data transfer nodes. This allows researchers/users and higher level meta-schedulers to use data placement as a service where they can plan ahead and reserve time and resources for their data movement operations. We present a novel approach for evaluating time-dependent structures with bandwidth guaranteed paths. We present a practical online scheduling model using advance reservation in dynamic network with time constraints. In addition, we report a new polynomial algorithm presenting possible reservation options and alternatives for earliest completion and shortest transfer duration. We enhance the advance network reservation system by extending the underlying mechanism to provide a new service in which users submit their constraints and the system suggests possible reservation requests satisfying users' requirements. We have studied scheduling data transfer operation with resource and time conflicts. We have developed a new scheduling methodology considering resource allocation in client sites and bandwidth allocation on network link connecting resources. Some other major contributions of our study include enhanced reliability, adaptability, and performance optimization of distributed data placement tasks. While designing this new data scheduling architecture, we also developed other important methodologies such as early error detection, failure awareness, job aggregation, and dynamic adaptation of distributed data placement tasks. The adaptive tuning includes dynamically setting data transfer parameters and controlling utilization of available network capacity. Our research aims to provide a middleware to improve the data bottleneck in high performance computing systems.



Document Availability at the Time of Submission

Release the entire work immediately for access worldwide.

Committee Chair

Kosar, Tevfik