In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

Olga Datskova

will defend her dissertation proposal

Prescriptive Analytics Engine for Unstable Events in Distributed Environments

Abstract

With projected growth of smart devices and network coverage, the notion of distributed environments is now extending beyond Grid and Cloud computing. In the context of naturally occurring events as well as changes in both security and governance, technological disruptions are inevitable in such systems. The need for effective tools in managing performance of network transfers, data distribution, application efficiency is evident. Lasting solutions stem from frameworks that are adaptable, scalab le and more importantly reach beyond predictive analytics.

Historical monitoring data for distributed environments provide key insights into real-world production environments as well as network behavior for data centers that spread across geographical distances. With constant data collection, processing and analysis of monitoring information, performance fluctuations can be detected and incorporated into disruptive event profiles. Characterization of these unstable regions provides the necessary intuition about building action cost analysis framework, which relies on adapting to available parameter manipulation within the locality of a data center, the user or central services, rather than acting on preset policies.

The goal of this research is to develop an engine that provides central services, data centers and users with best course of action during an ongoing instability occurring within a distributed system. Our aim is to uncover the nature of these unstable events, characterize environment-independent effects and provide a method for incorporating the developed decision framework into live production systems. For this research we are working with WLCG (Wordwide Large Hadron Collider Grid) Cloud services and ALICE (A Large Ion Collider Experiment) Grid environment that consists of over a 100 operational data centers in total spread across 42 countries.

Date: Tuesday, December 15, 2015
Time: 2:00 PM
Place: PGH 501D
Advisor: Dr. Larry Shi

Faculty, students, and the general public are invited.