DMCA
Complete Computer System Simulation: The SimOS Approach (1995)
| Venue: | IEEE Parallel and Distributed Technology |
| Citations: | 191 - 2 self |
BibTeX
@ARTICLE{Rosenblum95completecomputer,
author = {Mendel Rosenblum and Stephen A. Herrod and Emmett Witchel and Anoop Gupta},
title = {Complete Computer System Simulation: The SimOS Approach},
journal = {IEEE Parallel and Distributed Technology},
year = {1995},
volume = {3},
pages = {34--43}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
this article in fall 1993 and have been using it since early 1994. A simulation environment is only as good as the results it produces, and SimOS has proven to be extremely useful in our research. Recent studies in three areas illustrate the SimOS environment's effectiveness: Architectural evaluation: SimOS is playing a large part in the design of Stanford's Flash, a large-scale NUMA multiprocessor. 8 Researchers have used SimOS's detailed CPU modes to examine the per- formance impact of several design decisions. The ability to boot and run realistic workloads such as the Sybase database and to switch to highly accu- rate machine simulation have been of great value. System software development: In the design of an operating system for Flash, SimOS's direct-execution and binary-translation modes provide a development and debugging environment. Because SimOS supports full source-level debugging, it is a significanfly better debugging environment than the raw hardware. Developers use the detailed CPU mod- els to examine and measure time-critical parts of the software. In addition, SimOS provides the OS development group with Flash "hardware" long before the machine is to be complete. Workload characterization: SimOS's ability to run complex, realistic workloads including commercial applications enables researchers to characterize workloads that have not been widely studied before. Examples include parallel compilations and the Sybase database. 9 ased on our experience with SimOS, we believe that m'o of its features will become requirements for future simulation environments. These key features are the ability to model complex workloads, including all operating system activity, and the ability to dynam- ically adjust the level of simulation detail. Modern computer appl...
Keyphrases
simos approach complete computer system simulation sybase database realistic workload simulation environment flash hardware several design decision simos environment o development group architectural evaluation cpu mode workload characterization operating system activity time-critical part simulation detail system software development key feature modern computer appl accu rate machine simulation raw hardware large-scale numa multiprocessor detailed cpu mod el future simulation environment binary-translation mode debugging environment full source-level debugging large part recent study operating system per formance impact parallel compilation commercial application great value complex workload
