S.A. Herrod, Using Complete Machine Simulation to Understand Computer System Behavior, Ph.D. Thesis, Stanford University, http://simos.stanford.edu/papers.html, February 1998.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....is a well established technique for studying computer hardware and predicting system performance. Over the last years, many simulation systems with the goal of providing a general tool for such studies have been developed. Prominent systems include the full system wide simulation tools, like SimOS [5], SIMICS [9] and the SimpleScalar [1] an infrastructure for simulation and architectural modeling. These larger multiprocessor simulators model the computer hardware in high detail, allowing to evaluate the complete parallel architectures. SIMT, however, aims at supporting the study of ....

S. A. Herrod. Using Complete Machine Simulation to Understand Computer System Behavior. PhD thesis, Stanford University, February 1998.

....but also the prediction of new architectures. This approach is hence used within the work presented here. The simulation approach has long been used for predicting and analyzing performance on target architectures. Among the developed simulation tools, the most comprehensive system is SimOS [4], which models the entire uniprocessor and multiprocessor computer systems allowing a collection of performance data on different hardware components. The data collection is done by annotations which are simple Tcl scripts executed when an event of interest occurs in the simulator. Another well ....

S. A. Herrod. Using Complete Machine Simulation to Understand Computer System Behavior. PhD thesis, Stanford University, February 1998.

....of I O requires some similar techniques to the simulation of signals. Neither RSIM nor L RSIM support dynamic code generation as they predecode programs before simulation. Furthermore, although L RSIM simulates I O, it is inadequate for our requirement of simulating signals. SimOS SimOS [13, 17, 12] simulates hardware in sufficient detail to run a complete commercial operating system. We explored SimOSPPC, a version of SimOS developed by IBM which simulates the PowerPC processor and runs AIX. SimOS simulates both signal handling and dynamic code generation and therefore satisfies the ....

S. A. Herrod. Using Complete Machine Simulation to Understand Computer System Behavior. PhD thesis, Dept. of Computer Science, Stanford University, 1998.

....it comes to workloads that exhibit a significant amount of operating system activity, multiprogramming or I O, the assumptions made when such simulators were developed no longer hold, and a different approach must be taken. To address these workloads, several full system simulators such as SimOS [48] and SimICS [61] have been developed. These tools model I O devices to such detail that an almost unmodified operating system can be booted in the simulation environment, which allows researchers to evaluate virtually any workload. However, to achieve acceptable simulation performance, the ....

S.A. Herrod, Using Complete Machine Simulation to Understand Computer System Behavior, doctoral dissertation, Computer Science Dept., Stanford University, Stanford, Calif., 1998.

....same tiling. The computational overhead of tiling for a uniprocessor software renderer was measured by rendering the test scenes through our research rendering system Argus. In order to gather accurate, fine grained timing information, Argus was run on top of the full machine simulator SimOS [5]. The instrumentation capabilities of SimOS allowed measurements to be taken at a per pixel granularity without altering timing behavior. Figure 7 shows the ratio of rasterization time with tiling to that without tiling for each of the test scenes, as measured in Argus. The rasterization time is ....

S. Herrod. Using Complete Machine Simulation to Understand Computer System Behavior, Ph.D. thesis, Stanford University, February 1998.

....commodity desktop and server computer systems, as the simulators have been too slow. Recent advances in simulation technology have resulted in complete system simulators, modelling an entire workstation with sufficient performance and detail to run unmodified operating systems and large workloads [2, 3]. In this paper, we present a temporal debugger for soft real time Unix applications. It is based on the GNU debugger [1] modified to debug Linux applications running in a complete system simulator. The debugger is able to nonintrusively debug applications in the simulated system and present ....

S. A. Herrod. Using Complete Machine Simulation to Understand Computer System Behavior. PhD thesis, Stanford University, Feb. 1998.

....modern UNIX operating systems support, such as signal handlers, memory management, process creation and so on. In addition, resources used to store the cryptographic hashes that the XOM machine needs to ensure data integrity, must also be managed. To explore these matters further we used the SimOS [6] simulation system to model a XOM processor. We modified a SimOS model of an inorder processor using the MIPS R10000 [5] instruction set to behave as a XOM processor. The IRIX 6.5 [14] operating system from SGI was then modified to boot and run XOM applications in SimOS. With the experiments ....

S. A. Herrod. Using Complete Machine Simulation to Understand Computer System Behavior. PhD thesis, Stanford University, February 1998.

....are infeasible for simulating desktop systems, as simulation would be too slow. However, advances in simulation technology have resulted in simulators providing an approximate, but reasonably accurate timing model while executing with a slowdown of roughly 50 200 in relation to native execution [7, 10]. These sim ulators, referred to as complete system simulators, model an entire workstation at the instruction set level, and runs unmodified operating systems and workloads. A complete system simulator that is deterministic addresses the two major problems in realtime analysis: lack of ....

....during debugging. However, this prediction does not take operating system effects into account and works best for small programs. The work presented here is made possible by advances in simulation of computer systems, allowing the construction of accurate simulators of com plete computer systems [4, 7, 10]. The results pre sented in this paper have previously been presented in [1] which also contains further discussion on re lated work and issues important to debugging realtime programs. 3 Complete system simulation Many design and research areas benefit from simulation of computer systems. ....

Stephen Alan Herrod. Using Complete Ma- chine Simulation to Understand Computer System Behavior. PhD thesis, Stanford University, February 1998.

....of Java programs by examining both user and kernel execution. 3. Benchmarks and Experimental Methodology This work is based on simulation analysis of branch instruction traces generated on a complete system simulation environment. The experimental platform used to perform this study is SimOS [Rose95a, Herr97 and Herr98], a complete simulation environment that models hardware components with enough detail to boot and run a full blown commercial operating system. In this study, the SimOS version that runs the Silicon Graphics IRIX5.3 operating system has been used. The interpreter and JIT compiler of the Java ....

S. A. Herrod, Using Complete Machine Simulation to Understand Computer System Behavior, Ph.D. Thesis, Stanford University, Feb. 1998

....system performance in early stages of the development cycle, i.e. in predictive studies. Fig. 1. Replacing a complex model with a LDS When the system workload is transactional, it is advisable to represent the operating system and other overhead sources in the models. As it is remarked in [6], applications such as database management systems have significant operating system activity. The inability to include operating system overhead in models that represent transactional systems compromises their accuracy, and prevents a properly understanding of the workload behaviour. In the LDS ....

Herrod, S. A.: Using Complete Machine Simulation to Understand Computer System Behavior. PhD. Thesis. Stanford University (1998).

....However, the precise timing characteristics of a program may change from one run to the next, forcing the analyst to pursue a moving target. In recent years, complete machine simulation has become an increasingly important tool for the study of parallel systems performance and behavior. The SimOS [8] simulation environment has been used in a range of studies, and it has proven to be an effective tool for the analysis of computer systems. Several features of complete machine simulation are particularly valuable for the study of parallel applications: Visibility. SimOS provides complete ....

....simulation is significantly slower than regular program execution, simulation is most applicable to the study of several seconds or minutes of execution, not several hours or days. This problem can often be overcome by using faster, less detailed simulation modes to reach the section of interest [8]. Another concern with simulation is that most simulators do not fully model the detailed behavior of real hardware. Typically, if a problem is observed in the simulator it also occurs on the hardware, but the converse is not necessarily true. In the case of Argus, after fixing the problems found ....

S. Herrod. "Using Complete Machine Simulation to Understand Computer System Behavior." Ph.D. Thesis, Stanford University, February 1998.

....to cope with this complexity, system designers have developed new tools which are capable of capturing the behavior of these systems in great amounts of detail with minimal intrusiveness. Examples of these tools include: Complete machine simulation. Complete machine simulators such as SimOS [27] model all of the hardware of a computer system in enough detail to boot and run a commercial operating 1 CHAPTER 1. INTRODUCTION 2 system. The simulator provides total, non intrusive access to the complete hardware and software state of the system under simulation, from the contents of memory ....

....have begun to create comprehensive data collection tools for monitoring the performance and behavior of these computer systems. SGI s Performance CoPilot (PCP) 53] can collect data about all aspects of a system and report this data in real time; similarly, the SimOS complete machine simulator [27] can simulate the performance of an entire 52 CHAPTER 6. REAL TIME MONITORING: THE VISIBLE COMPUTER 53 Cluster 12 Clients Server Ethernet CPU Memory 2 Disks Network I F Bus 16 Nodes 10 Disks Network I F Interconnect R4K Core ICache DCache SCache TLB 2 CPUs Memory Controller ....

[Article contains additional citation context not shown here]

Stephen A. Herrod. Using Complete Machine Simulation to Understand Computer Systems Behavior. PhD thesis, Stanford University, February 1998.

....described by Bedichek [Bed90] and Magnusson et al. Mag93, MW95, Mag97] It executes at an approximate slowdown of 100. The SimOS group at Stanford University has built a machine simulator providing enough detail to run operating systems, requiring only small operating system modi cations [Her98] The simulator provides detailed instrumentation as well as support for programming event lters. It is also possible to program detailed timing analysis by associating scripts with events. The simulator supports switching between multiple timing models, thereby trading simulation performance ....

Stephen Alan Herrod. Using Complete Machine Simulation to Understand Computer System Behavior. PhD thesis, Stanford University, February 1998.

....encountered a scalability problem: Argus only scaled well to 26 processors before showing a sharp performance decline. They were unable to solve the problem using several traditional analysis tools, such as software profiling and hardware performance counters. We used Rivet coupled with SimOS [13], a complete machine simulator, to perform multiple simulation and visualization iterations, each focusing on different aspects of the application and operating system. Several types of per process events, such as thread scheduling and kernel traps, were displayed using two different metaphors ....

S. Herrod. "Using Complete Machine Simulation to Understand Computer System Behavior." Ph.D. Thesis, Stanford University, February 1998.

....a MIPS R10000 like [36] superscalar architecture. The overall design of the energy simulator is given in figure 1. We modified MXS CPU and the memory subsystem simulators to instrument accesses to their different components. This enables us to analyze our simulations using the Timing Trees [12] mechanism provided by SimOS. The MXS CPU simulator does not report detailed statistics about the memory subsystem behavior. Due to this limitation in SimOS, we use Mipsy for obtaining this information. 0000000 0000000 0000000 0000000 1111111 1111111 1111111 1111111 TCL Annotations ....

S. A. Herrod. Using Complete Machine Simulation to Understand Computer System Behavior. PhD thesis, Stanford University, February 1998.

....first approach that was used, and is, perhaps, the easiest to implement. In contrast, JIT compilers [2, 3, 4] which represent the state of the art, translate the byte codes to machine native code at runtime (using sophisticated techniques) for direct execution. It is becoming increasingly clear [7, 8, 9, 10] that accurate performance analysis requires an examination of complete system architecture and operating system behavior. While complete system simulation has been used to study several workloads [7, 8, 9] it has not been used in the context of Java programs or JVM implementations. A JVM ....

....sophisticated techniques) for direct execution. It is becoming increasingly clear [7, 8, 9, 10] that accurate performance analysis requires an examination of complete system architecture and operating system behavior. While complete system simulation has been used to study several workloads [7, 8, 9], it has not been used in the context of Java programs or JVM implementations. A JVM environment can be significantly different from that required to support traditional C or FORTRAN based code. The major differences are due to: 1) object oriented execution with frequent use of virtual method ....

[Article contains additional citation context not shown here]

S. A. Herrod, Using Complete Machine Simulation to Understand Computer System Behavior, Ph.D. Thesis, Stanford University, Feb. 1998.

....allowing it to run unmodified commercial operating systems 1 To be presented at MASCOTS 2000 and workloads. Advances in implementation technology allow such simulators to provide a reasonably accurate timing model while executing with a slowdown of roughly 50 200 in relation to native execution [8, 15]. Furthermore, complete system simulators can be designed to be fully deterministic. They thereby effectively address the two major problems in real time analysis: lack of repeatability and time distortion resulting from intrusion. Also, since simulators are implemented fully in software, they ....

....class of simulators provides a model of a machine at the instruction set level, as it represents a welldefined border between hardware and software. Variation of modeling an entire computer system at this level is a traditional theme. This type of simulation has been called complete machine [8], complete computer system [22] system level [12] instruction set [2] and faithful [5] simulation. In early work, it was referred to as virtual machines [4] or simply simulation [11] In this paper we shall use the term complete system instruction set simulation , or complete ....

[Article contains additional citation context not shown here]

S. A. Herrod. Using Complete Machine Simulation to Understand Computer System Behavior. PhD thesis, Stanford University, Feb. 1998.

....its implementation. To our knowledge, existing studies [1, 2, 3, 4, 5, 6, 7, 8] have been confined to examining JVM profiles from the architectural perspective, and there has been no attempt at understanding the influence of the operating system activities. It is becoming increasingly clear [9, 10, 11, 12] that accurate performance analysis requires an examination of complete system architecture and operating system behavior. Adhering to this philosophy, this paper presents results from an in depth look at complete system profiling of the SPECjvm98 benchmarks, focusing on the operating system ....

....part of sophisticated JIT compilers [19] Further, interpreters need a lower amount of memory than their JIT compiler counterparts, which can become important in resource constrained environments, such as hand held devices. While complete system simulation has been used to study several workloads [9, 10, 11], it has not been used in the context of Java programs or JVM implementations. A JVM environment can be significantly different from that required to support traditional C or FORTRAN based code. The major differences are due to: 1) object oriented execution with frequent use of virtual method ....

[Article contains additional citation context not shown here]

S. A. Herrod, Using Complete Machine Simulation to Understand Computer System Behavior, Ph.D. Thesis, Stanford University, Feb. 1998.

....at various points in time during the execution of the workload. We hand picked critical workload execution portions (for instance, remote process forks, or times with a high inter cell traffic) and concentrated most of our experiments on these parts. Using the annotation support provided by SimOS [Herrod98] we were able to automate this task by using just a few hundred lines of Tcl scripts [Ousterhout94] 3.5.2 Fault containment effectiveness To gain confidence in the correctness of our implementation, we first ran a large batch of fault injection experiments without using the operating system. In ....

.... memory management, and CPU scheduling) In the end we were left with about 3000 lines of C and assembly that are shared between the two implementations (see Table4.2 for a size comparison) The Cellular Disco prototype was developed entirely using the SimOS system level simulator [Rosenblum95a][Herrod98]. We relied heavily on non intrusive SimOS annotations written in the scripting language Tcl [Ousterhout94] to check various assertions about the correctness of the Cellular Disco code. SimOS annotations are superior to regular assertions that are part of the code since there are no restrictions ....

S. Herrod. "Using Complete Machine Simulation to Understand Computer System Behavior." Ph.D. Thesis. Stanford University Technical Report CS-TR-98-1603, February 1998.

No context found.

S.A. Herrod, Using Complete Machine Simulation to Understand Computer System Behavior, Ph.D. Thesis, Stanford University, http://simos.stanford.edu/papers.html, February 1998.

No context found.

S.A. Herrod, Using Complete Machine Simulation to Understand Computer System Behavior, Ph.D. Thesis, Stanford University, http://simos.stanford.edu/papers.html, February 1998.

No context found.

Stephen Alan Herrod. Using Complete Machine Simulation to Understand Computer System Behavior. PhD thesis, Stanford University, February 1998.

No context found.

Herrod, S.A., "Using Complete Machine Simulation to Understand Computer System Behaviour," Doctoral Dissertation, Feb. 1998.

No context found.

Stephen A. Herrod. "Using Complete Machine Simulation to Understand Computer System Behavior." Ph.D. Thesis, Stanford University, February 1998.

No context found.

S. A. Herrod. Using Complete Machine Simulation to Understand Computer System Behavior. PhD thesis, Stanford University, Feb. 1998. 3, 3.4

No context found.

S. A. Herrod, "Using Complete Machine Simulation to Understand Computer System Behavior," Ph.D. thesis, Stanford University, Feb. 1998.

No context found.

S. A. Herrod. Using Complete Machine Simulation to Understand Computer System Behavior. PhD thesis, Stanford University, Feb. 1998.

No context found.

S. A. Herrod. Using Complete Machine Simulation to Understand Computer System Behavior. PhD thesis, Stanford University, February 1988.

No context found.

Steve Alan Herrod. Using Complete Machine Simulation to Understand Computer System Behavior. PhD thesis, Stanford University, 1998.

No context found.

S. A. Herrod. Using Complete Machine Simulation to Understand Computer System Behavior. PhD thesis, Stanford University, February 1998.

No context found.

Stephen Herrod. Using Complete Machine Simulation to Understand Computer System Behavior. PhD thesis, Stanford University, USA, February 1998.

No context found.

Herrod, S. A. 1998a. Using Complete Machine Simulation to Understand Computer System Behavior. February. PhD Thesis, Department of Computer Science, Stanford University.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC