|
CMU-CS-00-100
Computer Science Department
School of Computer Science, Carnegie Mellon University
CMU-CS-00-100
Modeling and Scheduling of MEMS-Based Storage Devices
John Linwood Griffin, Steven W. Schlosser, Gregory R. Ganger, David F. Nagle
November 1999
Keywords: MEMS, simulation, request scheduling
MEMS-based storage devices are seen by many as promising replacements for
disk drives. Fabricated on CMOS, MEMS-based storage uses thousands of
small, mechanical probe tips to access Gigabytes of nonvolatile storage.
This paper takes a first step towards understanding the performance
characteristics of these devices. Using trace-driven simulation and models
based on the physical equations that govern the device's basic
characteristics, this work explores how different physical characteristics
(e.g., acceleration, data rates) and scheduling algorithms impact the
design and performance of MEMS-based storage. Our results show that
MEMS-based storage can improve storage access rates by a factor of 5 over
conventional disk-based storage, with average access times of under 2 ms.
Further, our analysis of scheduling algorithms shows that the relative
benefits of request scheduling are similar to standard disks.
22 pages
Superceded by:...
Modeling and Performance of MEMS-based Storage Devices
John Linwood Griffin, Steven W. Schlosser, Gregory R. Ganger, David F. Nagle In Proceedings ACM SIGMENTRICS 2000,
International Conference on Measurement and Modeling of Computer Systems, June 2000, pp 56:65.
MEMS-based storage devices are seen by many as promising alternatives to
disk drives. Fabricated using conventional CMOS processes, MEMS-based
storage consists of thousands of small, mechanical probe tips that access
gigabytes of high-density, nonvolatile magnetic storage. This paper takes
a first step towards understanding the performance characteristics of
these devices by mapping them onto a disk-like metaphor. Using simulation
models based on the mechanics equations governing the devices' operation,
this work explores how different physical characteristics (e.g., actuator
forces and per-tip data rates) impact the design trade-offs and
performance of MEMS-based storage. Overall results indicate that average
access times for MEMS-based storage are 6.5 times faster than for a modern
disk (1.5 ms vs. 9.7 ms). Results from filesystem and database benchmarks
show that this improvement reduces application I/O stall times up to 70%,
resulting in overall performance improvements of 3X.
and.....
Operating System Management of MEMS-based Storage Devices
John Linwood Griffin, Steven W. Schlosser, Gregory R. Ganger, David F. Nagle
In 4th USENIX Symposium on Operating Systems Design and Implementation
(OSDI), October 2000.
MEMS-based storage devices promise significant performance, reliability,
and power improvements relative to disk drives. This paper explores how
the physical characteristics of these devices change four aspects of
operating system management: request scheduling, data placement, failure
management, and power management. Adaptations of disk request scheduling
algorithms are found to be appropriate for these devices; however, new
data placement schemes are shown to better match their differing
mechanical positioning characteristics. With aggressive internal
redundancy, MEMS-based storage devices can tolerate failure modes that
cause data loss for disks. In addition, MEMS-based storage devices enable
a finer granularity of OS-level power management because the devices can
be stopped and started rapidly and their mechanical components can be
individually enabled or disabled to reduce power consumption.
|