D. A. Henke. Loki -- An empirical evaluation tool for distributed systems: The experiment analysis framework. Master's thesis, University of Illinois at Urbana-Champaign, 1998. Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
An Experimental Evaluation Of The Coda Distributed File System.. - Lefever (2003) (1 citation) (Correct)
....phase. It converts the local times of recorded experiment data into a single global timeline. The analyzer also checks the fault injections for correctness using the global timeline. The details of o#ine analysis and formulation of global timelines are not reviewed here. They can be found in [14, 18]. Following the analysis phase, the Loki measure estimator uses the global timelines from the experiments with correct fault injections to estimate user specified measures in the measure estimation phase. From the measure estimation phase, experiments containing incorrect fault injections can be ....
....execution, the campaign is ready to be processed offline by Loki. As described earlier, there are two phases to the o#ine processing: experiment analysis and measure computation. Each step is invoked from the Loki Manager. Experiment analysis performs the analysis phase of Loki, as described in [14, 18], to create global timelines and determine the correctness of fault injections from the raw data collected during the campaign s execution. The Experiment Analysis Window, pictured in Figure 2.8(b) displays several details of an analyzed experiment, such as the # and # values for the clock ....
D. A. Henke, "Loki -- an empirical evaluation tool for distributed systems: The experiment analysis framework," M.S. thesis, University of Illinois at Urbana-Champaign.
Evaluating Unavailability Caused By Group Membership Using.. - Joshi (2003) (1 citation) (Correct)
....point to note is that local events are measured in terms of local times on the hosts on which they occur. Converting these local times into a global time as required for the global timeline represents a significant challenge. Loki utilizes an o#ine clock synchronization algorithm (described in [Hen98] to perform this conversion. This clock synchronization algorithm can place deterministic global time bounds on a local event occurrence. This means that the global timelines that are actually generated are not ideal because they contain sequences of event occurrence intervals instead of ....
....are combined to form 9 (Reference Machine) Local Timelines Global Timeline Offline Clock Synchronization Node1 Node2 Node3 Figure 2.1: Global Timeline Generation a global timeline for an experiment execution. The lengths of these intervals depend on several factors (which are described in [Hen98] and are quantified for a typical case in Section 2.4. However, there always exists a reference host (usually the fastest machine) whose clock is used as the global time. The precise timing for events occurring on this host is known, and the length of the event intervals for this host is zero. ....
[Article contains additional citation context not shown here]
D. A. Henke. Loki -- An empirical evaluation tool for distributed systems: The experiment analysis framework. Master's thesis, University of Illinois at UrbanaChampaign, 1998.
Loki - An Empirical Evaluation Tool For Distributed Systems: The.. - Pistole (1998) (Correct)
....and architecture employed in building the system, the components in a model as well as many of the relationships between these 2 components can be defined. Some parameters, however, must be either estimated or determined through some external source. Loki, described in this thesis and in [2], is intended mainly for the purpose of obtaining some of these parameters for fault tolerant software distributed systems; however, it can also be used for more general computer systems. The values to be obtained for these systems are related to how often and how quickly the fault tolerant system ....
....input for system models that are meant to predict behavior in system configurations that are difficult to create using a testbed (e.g. systems with a very large number of hosts) Accomplishing this second goal requires several steps that are discussed briefly in Section 2. 4 and in more detail in [2]. For the Loki empirical evaluation tool, these goals were addressed with the additional constraint that it must be possible to test systems using multiple hosts on a testbed. The alternative, simulation of multiple networked hosts on a single computer, requires specialized 6 real time operating ....
[Article contains additional citation context not shown here]
D. Henke, "Loki -- an empirical evaluation tool for distributed systems: The experiment analysis framework," M.S. thesis, University of Illinois, Urbana, IL, 1998.
A Global-State-Triggered Fault Injector for.. - Chandra, +, +.. (2002) (Correct)
No context found.
D. A. Henke. Loki -- An empirical evaluation tool for distributed systems: The experiment analysis framework. Master's thesis, University of Illinois at Urbana-Champaign, 1998.
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