Advancement: Predicting the Future of Archival Storage Using Simulation

Speaker Name: 
James Byron
Speaker Title: 
PhD Student (Advisor: Ethan Miller)
Speaker Organization: 
Computer Science
Start Time: 
Thursday, December 13, 2018 - 10:00am
End Time: 
Thursday, December 13, 2018 - 12:00pm
Engineering 2, Room 380
Ethan Miller

Abstract:  Archival storage systems utilize large numbers of storage devices to reliably store increasing amounts of data. Today, numerous storage technologies are available for use in archival storage systems, including tape, optical disc, hard disk, and solid state disk. In order to satisfy an ever-increasing demand for archival storage capacity and performance, numerous prospective technologies have been proposed that may offer greater capacity or performance than today's storage technologies for archival storage systems. Such prospective technologies include synthetic DNA, glass, and carbon nanotubes. While the promise of prospective archival storage technologies is to increase storage capacity, their cost and performance must be sufficient to compete with existing technologies and to meet the unique demands of each archive's workload. What values of capacity, cost, performance, and reliability must prospective storage technologies offer in order to become economically viable for archival workloads, and in what ways must existing technologies develop to meet the demands of future archival storage systems? How will specific workload requirements affect the cost of using each storage technology in an archive?

I propose a simulator that will model existing and prospective archival storage technologies under varying assumptions about their development rates in order to compare their cost and performance in archival storage systems. The simulator will also define lower bounds of reliability and the rates of development for cost-per-byte and performance that both existing and prospective storage technologies must achieve in order to meet future demands for archival storage. Finally, my simulations will include parameters to define the workload of the archival system to model how much it will cost to use each storage technology in an archival system with a specific workload.