This project addresses the use of flash memory as the foundation for multi-tiered storage systems. Digital content (e.g., text, image, sound, video, etc.) has experienced explosive growth. Internet media providers (e.g., Facebook, Flicker, YouTube, Hulu, Netflix, etc.) are facing huge I/O demand for delivering content. And, due to the increasing video traffic on the web, backend storage servers not only need increased storage capacity but also increased storage bandwidth to serve the web clients in a timely manner.

Hard disk drives (HDDs) have advanced in capacity per dollar while main memory (in the form of DRAM) has advanced in speed per dollar. Recently, flash memory technology has been getting attention from industry due to its significant cost decrease. Flash-based Solid-State Drives (SSDs) are now at the price point where they can bridge the capacity-speed gap between HDDs and DRAM. In addition, energy consumption of flash memory is far less than the two traditional storage technologies. This project proposes that multi-tiered storage architecture is a compelling vehicle to optimize capacity, speed, and energy consumption of storage by exploiting the advantages of all three technologies

Storage devices make up nearly 40% of the energy consumption of a typical data center -- consequently, storage servers should not only have high capacity and deliver high performance, but also be energy efficient. From an environmental perspective, it is extremely important to devise solutions that can reduce the carbon footprint of data centers. A multi-tiered, web storage system that uses flash memory for caching is a promising solution for performance, capacity, and energy efficiency. This project's technology presents an opportunity for CDN providers and web service providers to optimize the performance and energy efficiency of their storage systems. While the primary customers of this technology will be corporations that run a large number of web servers, there is an environmental benefit from the carbon emission reduction of the data centers run by such big web corporations.

Project Start
Project End
Budget Start
2014-02-01
Budget End
2016-07-31
Support Year
Fiscal Year
2014
Total Cost
$50,000
Indirect Cost
Name
Georgia Tech Research Corporation
Department
Type
DUNS #
City
Atlanta
State
GA
Country
United States
Zip Code
30332