Filling the Memory Access Gap: A Case for On-Chip Magnetic Storage

Steven W. Schlosser, John Linwood Griffin David F. Nagle, Gregory R. Ganger

Dept. Electrical and Computer Engineering
Carnegie Mellon University
Pittsburgh, PA 15213
{schlos, griffin2, David.Nagle, ganger }@ece.cmu.edu
http://www.pdl.cs.cmu.edu/

Abstract

For decades, the memory hierarchy access gap has plagued computer architects with the RAM/disk gap widening to about 6 orders of magnitude in 1999. However, an exciting new storage technology based on MicroElectroMechanical Systems (MEMS) is poised to fill a large portion of this performance gap, delivering significant performance improvements and enabling many new types of applications. This research explores the impact MEMS-based storage will have on computer systems. Working closely with researchers building MEMS-based storage devices, we examine the performance impact of several design points. Results from five different applications show that MEMS-based storage can reduce application I/O stall times by 80-99%, with overall performance improvements ranging from 1.1X to 20X for these applications. Most of these improvements result from the fact that average access times for MEMS-based storage are 5 times faster than disks (e.g., 1-3ms). Others result from fundamental differences in the physical behavior of MEMS-based storage. Combined, these characteristics create numerous opportunities for restructuring the storage/memory hierarchy.

Click here for the full paper in pdf or postscript.