D. Taylor. A Prototype Debugger for Hermes. Cascon '92, IBM Canada Ltd, Toronto, pp. 29 - 42, November, 1992.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... several environments: a functional prototyping environment (MLOG [16] a commercial prototyping environment (ObjecTime [36] a distributed software system simulator called Parasol [34] coarse grained UNIX tasks [27] and in the DCE RPC environment [37] using data collected by the POET debugger [49]. We concentrate here on traces produced by a design prototype environment, namely ObjecTime. There are several benefits to using TLC as opposed to a source code examination approach for constructing models. Traces incorporate the dynamic details of a design that are difficult to determine from ....

D. J. Taylor. "A prototype debugger for Hermes." In CASCON 92, pages 29--42, 1992.

.... prototyping environment (MLOG [16] a commercial evolutionary prototyping environment (ObjecTime [38] a distributed software system simulator called Parasol [34] in coarse grained UNIX tasks [22] and in the DCE RPC multi threaded environment [39] using data collected by the POET debugger [49]. In these environments, proper time establishes a causal as well as a temporal (happened before) relation between events. Many uses for the temporal relation have already been established: distributed algorithm implementation [32] system feature implementation (e.g. causal memory [28] causal ....

....nested within the other: Task A initiates the RPC and blocks at event , and the nested blocking interaction is initiated by Task B at event . One approach to identifying task blocking is to augment the timestamps (i.e. metrication) with information about the communication mechanism, as is done in [49]. Other approaches have modified the happened before relation ( 8] 14] and [11] The topology of proper time directly characterizes task blocking by labeling blocking and non blocking events differently. Then the Figure 5: Realized, Visible, and Potential Causality Task B Task A e A1 e ....

D. J. Taylor. "A prototype debugger for Hermes." In CASCON 92, pages 29--42, 1992.

.... tool of ObjecTime [14] Schmidt s ACE toolkit [17, 16] an object oriented functional prototyping environment (MLOG [10] a distributed software system simulator called Parasol [13] coarse grained UNIX objects [9] and in the DCE RPC environment [15] using data collected by the POET debugger [20]. The ANGIOTRACE was introduced in [6] by the author and has been further described in [7, 8] 1. Angiotrace is a trademark of Angiograms for Software Analysis Inc. who can be contacted at angio istar.ca. ANGIOTRACE appears in capital letters to signify it is a trademark. V1.2 Page 3 22 July ....

D. J. Taylor. A prototype debugger for Hermes. In CASCON 92, pages 29--42, 1992.

....essential for a deterministic replay is collected, minimizing interference with the non deterministic program behavior. Deterministic replays will then be used to analyze program behavior. To overcome the need for determining the global state of a distributed computation, many researchers [3, 8, 20, 21] have chosen an event based approach. Event information can be captured relatively easily compared to state information. In addition, states of a computation can always be restored from event information. What is important in an event based approach to debugging is how events are causally related ....

....and restoring a meaningful state of a computation thus takes two replays. The strategy outlined here can be used with any kind of breakpoint detection algorithm. The research described in this paper is part of a project at the University of Waterloo to design and implement a distributed debugger [21]. The current prototype records event information for the purpose of visualization. It has tools for visualizing events and their causal relationships. It also has some tools for determining and visualizing abstract program behavior [13] Since most tools use vector time, much effort is put into ....

D.J. Taylor. A Prototype Debugger for Hermes. In Proceedings of the 1992 CAS Conference, Volume I, pages 29--42, Toronto, Ontario, Canada, November 1992. IBM Canada Ltd. Laboratory, Centre for Advanced Studies. A Appendix: Proofs Theorem 5.11. For any A ` E, pgs:A = T :pA inf (E \Gamma T R :fA) = egs:A inf (E \Gamma (SUP a : a 2 A : T

....for event collection, analysis and replay. We are aware of this requirement and have begun to work on this issue. C programs can already be compiled with event tracing for all accesses to coroutines, monitors, and tasks [2] The events are currently displayed by a powerful event tracing viewer [21] (see Figure 4) The next step is to connect KDB, the local debugger, and the event tracer to provide debugging support for deadlock and race conditions. KDB will be extended with a programmatic interface, so that its functionality can be accessed by commands sent through a socket instead of ....

Taylor, D. "A Prototype Debugger for Hermes". In Proceedings of the 1992 CAS Conference, volume 1, pages 29--42, Toronto, Ont., Canada, November 1992. IBM Canada Ltd. Laboratory, Centre for Advanced Studies.

.... prototyping environment (MLOG [16] a commercial evolutionary prototyping environment (ObjecTime [38] a distributed software system simulator called Parasol [34] in coarse grained UNIX tasks [22] and in the DCE RPC multi threaded environment [39] using data collected by the POET debugger [49]. In these environments, proper time establishes a causal as well as a temporal (happened before) relation between events. Many uses for the temporal relation have already been established: distributed algorithm implementation [32] system feature implementation (e.g. causal memory [28] causal ....

....nested within the other: Task A initiates the RPC and blocks at event , and the nested blocking interaction is initiated by Task B at event . One approach to identifying task blocking is to augment the timestamps (i.e. metrication) with information about the communication mechanism, as is done in [49]. Other approaches have modified the happened before relation ( 8] 14] and [11] The topology of proper time directly characterizes task blocking by labeling blocking and non blocking events differently. Then the causal chain of events through the RPC is immediately identifiable. Proper time ....

D. J. Taylor. "A prototype debugger for Hermes." In CASCON 92, pages 29--42, 1992.

....is caused in part by the lack of a total ordering in such executions [25] The use of tools which help understand the behaviour of distributed executions is steadily increasing. The most common of these are event visualisation tools, which provide the user with a graphical view of the execution [14, 15, 20, 28, 35, 41]. This view presents the occurrence of different types of events over time. An event is a conceptual entity that causes a change in the state of the execution. Each event occurs at the instant at which some predefined computation is completed. Such events can be the sending and receiving of ....

....allowing for what is known as real time displays [37] A fundamental difference between event visualisation tools and traditional sourcecode debuggers is that, unlike code statements, the events of an execution are not known a priori. Because of this limitation, many distributed debugging tools [14, 15, 20, 28, 35, 41] deal with post mortem debugging. In this type of approach, the application is run to completion, while data describing its behaviour is collected. The data can then repeatedly be used by the tool to describe the execution as it has previously occurred. This method allows the events making up the ....

[Article contains additional citation context not shown here]

D. J. Taylor. A prototype debugger for Hermes. In Proceedings of the 1992 CAS Conference, volume 1, pages 29--42, November 1992.

....similar to the difference between program development using language compilers and interpreters. Most PPSs use a post mortem strategy for two reasons; it is simpler to implement and it provides more user control. Event acquisition and analysis have been wellstudied in the literature (for example, [7]) The key factor in acquisition is to reduce the probeeffect, the perturbation of the performance of a program due to the execution of instrumentation code. This intrusion can potentially alter the execution of a program to the extent of altering the results. The Enterprise approach to the probe ....

....and other graphical displays not only grab attention, but are ideal in capturing the dynamic nature of parallel programs. Well designed graphical displays with simple visual cues and familiar symbolic representations enhance a user s innate ability to process visual information (for example, [7, 1]) 3 Despite the power of graphical visualization, there are several limitations to such presentation systems. First, if too much information is displayed at one time, or if displayed information is changed too rapidly, the user will miss fine details that may be of interest. Second, there are ....

[Article contains additional citation context not shown here]

D. Taylor. A Prototype Debugger for Hermes. CASCON '92, pp. 29-42, 1992.

....conclusions of the thesis and discusses future work. Chapter 2 Visualizing Object Oriented Executions Section 2.1 describes the model of object oriented concurrency investigated by this thesis, which uses two pieces of software to demonstrate some of the concepts it explores. The first is Poet [26], a partial order event tracer developed at the University of Waterloo. It uses process time diagrams to visualize distributed and concurrent program executions, and is described in Section 2.2. The second piece of software is ABC (Active Base Class) 20] a class library for adding ....

David Taylor. A prototype debugger for hermes. In Proceedings of IBM CASCON, pages 29--42, Toronto, Ontario, November 1992.

....facilities for event capture or recording time differences. PVM [8] provides a special service layer (like PICL) and distributed control but lacks event collection. MIDAS [1] provides control and instrumentation but only runs in a simulated environment, rather than on a real network. POET [9, 10] provides event collection but not experiment control. In fact, none of the other tools provides such complete control of an experiment as DECALS, as listed above. The next section outlines the main features of DECALS, then its structure and operation are briefly described. The range of potential ....

David J. Taylor. A prototype debugger for hermes. In CASCON 92, pages 29--42, 1992.

....interactions and the network communication, which could provide valuable information for debugging distributed programs. A powerful debugging tool is urgently needed for this new type of distributed programming environment. This essay examines how an existing event based debugger, Poet [3] (Partial Order Event Tracer) can be adapted to visualize Java internal synchronization events as well as the external communication activities. A brief introduction to event based debugging is presented in the following section, followed by a discussion of the shortcomings of applying the ....

....execution to high level network communication. Instrumentation will introduce more or less of a probe effect into an execution: the program behaviour may change when probes are inserted to collect information. CHAPTER 1. INTRODUCTION 4 Figure 1.1: Poet sample visualization 1. 2 Poet Poet [3], a partial order event tracer, was developed by Prof. D.J. Taylor at the University of Waterloo. The tracer uses process time diagrams to present the event data gathered from instrumented program execution graphically. It has additional features available to assist users in visualization, such as ....

David J. Taylor. A prototype debugger for Hermes. In Proceedings of the 1992 CAS Conference, pages pp. 29--42, November, 1992.

....are discussed. The implementation of the facilitators is based on extending POET to include debugging functions. POET was developed by David Taylor at the University of Waterloo and can be best described as a tool for instrumenting a distributed application to collect event traces of its execution [Tayl92, Tayl93]. Using POET as the foundation, the implementation described in this thesis has added distributed breakpoint and execution based replay facilities which collectively make up the facilitator manager. 5.1 Replay Facilitator The behaviour of a distribution application is non deterministic because of ....

D.J. Taylor, `A prototype debugger for Hermes', Proceedings of the 1992 CAS Conference, vol. I, pp. 29-42, November 1992.

....might be specified as follows: f1g : user. f1g , block , 5sec: Gamma [ f1g , unblock ] Chapter 4 Implementation 4.1 POET 4.1. 1 Motivation The Partial Order Event Tracer (POET) is an event visualization tool developed by the Shoshin research group at the University of Waterloo [38, 40]. There are several reasons why POET provides a good base for event detection investigations: ffl Solid (tested) foundation. The POET system has been in use for four years, and the algorithms to timestamp events and display process time diagrams are reasonably fast and largely bug free. ffl ....

....so that practical experience with it can shed further light on the ideal form for such a tool; POET seems to have the user base that is a prerequisite for such an iterative refinement process. 4.1. 2 Architecture POET is target system independent [41] and has been adapted for use with Hermes [38], OSF DCE [43] PVM [25] ABC [33] C and, naturally, to debug itself. Various custom parameters for each of these environments are stored in targetdescription files; the relevant environment is identified when the first debuggee starts, and the event tracer then configures itself ....

[Article contains additional citation context not shown here]

David J. Taylor. A prototype debugger for Hermes. In Proceedings of the 1992 CAS Conference, volume 1, pages 29--42, 1992.

....distributed environments. Hence, tools that help understand the behavior of such applications are becoming more and more common. A useful method for describing and reasoning about a distributed execution is the use of eventvisualization tools, which give the user a graphical view of the execution [5, 8, 12, 15, 18]. These tools aim to help the user understand and debug a distributed application. Although visualization is a very useful method for such a task, it is often not sufficient for full debugging purposes. Traditional debugging operations are also needed in such tools, so a user can study a faulty ....

....task. This paper describes a well defined method for single stepping in such a tool, which allows the user to better understand the behavior of the execution. The concepts explained here are currently implemented in Poet, a Partial Order Event Tracer, developed at the University of Waterloo [12, 13]. An example of Poet during the visualization of a PVM (Parallel Virtual Machine) 4] application is given in Figure 1. 1 Within Poet, at the simplest level, each entity exhibiting sequential behavior is represented by a horizontal trace line. These entities can be processes, tasks, threads, ....

D. J. Taylor. A prototype debugger for Hermes. In Proceedings of the 1992 CAS Conference, volume 1, pages 29--42, November 1992.

....interprocess interactions of the processes making up the distributed application. A graphical display showing the events causing interactions between processes (a process time diagram) can be very useful in understanding and debugging a distributed program. The Partial Order Event Tracer (POET) [21, 22, 23, 24], developed by Prof. David Taylor at the University of Waterloo, produces such a display with a full set of additional features and visual enhancements, such as independent event scrolling and predecessor successor coloring (see Figure 1.1) 1.2 Research Motivation Typically, the lowest level of ....

....debugger involves various design decisions. The following sections describe the alternatives considered and the reason for making each choice in designing the linkage. The reader is assumed to be familiar with the architecture (the standard set of clients and server) of POET, as described in [21]. 2.1.1 User Interface The event display generated by POET provides a high level view of the distributed program execution. This should be the level of interest to the user when debugging first starts. As debugging proceeds, the user may want a lower level view of the program execution; the ....

David J. Taylor. A prototype debugger for Hermes. Proceedings of the 1992 CAS Conference, Volume I, pages 29--42, November 9--12 1992. BIBLIOGRAPHY 57

....Even a moderate Hermes application consists of a large number of processes. Hermes therefore is a convenient vehicle for testing our ideas in a complex environment. A last, more pragmatic reason is that we had access to the source of an existing visualization tool for Hermes applications [16, 18]. A separate paper in these proceedings ( 18] discusses the basic visualization tool and presents one abstraction technique, process clustering. The next section introduces another abstraction technique, grouping events into primitive events. Next, two tools to derive suitable abstractions ....

David J. Taylor. A Prototype Debugger for Hermes. In Proceedings of the 1992 CAS Conference, Volume I, pages 29--42, Toronto, Ont., Canada, November 1992. IBM Canada Ltd. Laboratory, Centre for Advanced Studies.

....plan libraries, use the library to search for programming concepts in the source code, i.e. they perform a static analysis of the application [22, 23, 24] We search for patterns throughout the event data, produced by the execution of the application. Our analysis is therefore dynamic. We use Poet [31, 32], an event tracer developed at the University of Waterloo. Poet visualizes a distributed execution in order to provide the programmer with a better understanding of the distributed application. However, large distributed applications tend to produce a huge number of events, and a significant ....

....two timestamp tests correctly deduce the relation, using the above timestamps: ffl Timestamp Test 1: e f if and only if T e [i] T f [i] for all timestamp vector elements i. ffl Timestamp Test 2: e f if and only if T e [p] T f [p] where event e occurs in process p. 2. 2 Poet Poet [31, 32] is a partial order event tracer developed at the University of Waterloo. It uses process time diagrams to visualize distributed program executions. Processes are represented by horizontal lines, called traces, with time progressing from left to right. The visualization focuses on interactions ....

[Article contains additional citation context not shown here]

D.J. Taylor. A prototype debugger for hermes. In Proceedings of IBM CASCON, pages 29--42, Toronto, Canada, November 1992.

....To better understand the behavior of distributed applications, many tools and algorithms have been developed. To mention just a few of them, one could think of snapshot algorithms [10, 22, 25, 29] predicate detection algorithms [14] and debugging tools including tools for visualizing behavior [35], breakpoint facilities [20, 28, 31] and tools for recognizing behavioral patterns [5, 23] All these tools and algorithms have one thing in common: They use event based models of distributed computations. More in particular, they use the causal relationships among events to extract information ....

D.J. Taylor. A Prototype Debugger for Hermes. In Proceedings of the 1992 CAS Conference, Volume I, pages 29--42, Toronto, Ont., Canada, November 1992. IBM Canada Ltd. Laboratory, Centre for Advanced Studies.

....[22] Even a moderate Hermes application consists of a large number of processes. Hermes therefore is a convenient vehicle for testing our ideas in a complex environment. A last, more pragmatic reason is that we had access to the sources of an existing visualization tool for Hermes applications [61]. This paper is organized as follows. After this introduction, a visualization frequently used to depict distributed executions is shown. Next, two tools to derive suitable abstractions automatically are presented. Using a sample distributed application, different abstract visualizations of one ....

....compiler is invoked. This execution creates a total of 175 processes, and the event trace contains 2534 primitive events. Figure 8: Low level Visualization of the makehermes Execution Figure 8 depicts a segment of the execution history at the lowest abstraction level, using the tool described in [61, 62]. This tool draws a set of horizontal lines, one for each process, placing a symbol on the appropriate line for each event. Time flows from left to right, and a scrollbar allows for scrolling in the vertical (process) dimension. Scrolling in the time dimension is more complex since it depends on ....

David J. Taylor. A Prototype Debugger for Hermes. In Proceedings of the 1992 CAS Conference, Volume I, pages 29--42, Toronto, Ont., Canada, November 1992. IBM Canada Ltd. Laboratory, Centre for Advanced Studies.

....examples of the required tools. A third paper, Linking Specification, Abstraction and Debugging, describes a prototype tool for capturing and visualizing events among components of distributed applications. It also discusses these capabilities as integral components of a (distributed) debugger [Tay92] The remainder of this paper contains Sections 2 7. The following section defines some of the terminology used in this paper. Section 3 discusses the motivation for the CORDS project and describes our long term view of distributed application environments. Section 4 presents a brief overview of ....

....developer passes the output of the OSF DCE Interface Definition Language (IDL) compiler to a post processor. The post processor automatically instruments the RPC client and server to send communication event messages to a waiting event collector. The distributed debugger (Event Collector, Shoshin) Tay92] collects communication events from the above RPC tracer as well as the communication monitored by the process server. Events are displayed on event time lines representing running processes. The debugger could also pass information to other tools, such as Hy , for subsequent display and ....

D. Taylor. A prototype debugger for Hermes. In CASCON '92 Proceedings, volume I, pages 29--42. The IBM Toronto Centre for Advanced Studies, November 1992. Also in Volume II, pp. 313-326.

....distributed or parallel application involves all of the problems encountered in debugging a sequential program, plus problems specific to the distributed parallel environment. A useful facility in debugging such applications is a display showing the interactions between processes. A previous paper [9] described the value of such displays in detail and their implementation in a prototype debugger for the Hermes language [2] That paper also provided a brief description of process clustering as a means for removing currently unwanted detail from a display and the techniques adopted for effective ....

....and thus is likely to cause the user to identify the displayed event with the original event rather than imagining it to be a new interface event. There is still the problem of what action to take if concurrency exists among the events to be displayed for a cluster. As described previously [9], the approach adopted is to dynamically create enough trace lines to allow all mutually concurrent events of a cluster to be displayed on separate trace lines. Sections of these trace lines are then joined to each other by linking arrows to provide a visual indication of the precedence ....

[Article contains additional citation context not shown here]

D. J. Taylor, "A prototype debugger for Hermes," Proceedings of the 1992 CAS Conference, Volume I, pp. 29-42 (November 9-12, 1992).

....representing some activity performed by a process and considered to take place at an instant in time. Typically, the lowest level of observed behaviour consists of events representing process interactions, such as sending and receiving messages and process creation and termination. Our tool, Poet [11, 12], displays processes and events using two dimensional process time diagrams. The placement of events along the time axis is based on either their occurrence in real time or their relationship to other events in the partial order introduced by Lamport [9] Each display mode has value: for example, ....

David J. Taylor. A prototype debugger for Hermes. In Proceedings of the 1992 CAS Conference, Volume I, pages 29--42, Toronto, Ont., Canada, November 1992. IBM Canada Ltd. Laboratory, Centre for Advanced Studies.

No context found.

D. Taylor. A Prototype Debugger for Hermes. Cascon '92, IBM Canada Ltd, Toronto, pp. 29 - 42, November, 1992.

No context found.

D. Taylor, `A prototype debugger for Hermes', in Proceedings of the 1992 CAS Conference, Volume 1, Center for Advanced Studies, IBM Canada, Ltd, Laboratory, November 1992, pp. 29--42,.

No context found.

D. J. Taylor. A prototype debugger for Hermes. In CASCON 92, pages 29--42, 1992.

First 50 documents

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