Henrik Thane and Hans Hansson. Using deterministic replay for debugging of distributed real-time system. In Proceedings of the 12th euromicro conference on real-time systems, pages 256--272, Stockholm, June 2000. IEEE Computer Society, IEEE Computer Society Press.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....Thus, for example, a probe e ect can be visible if a software is modi ed, or if the platform of the system is changed. As a consequence of this, the perturbation from the probes must not change over time. Thus, probes should be left in the system even after the end of the veri cation process [6] [8, s3.1;p51] 1.2 The E ect of Jitter on Testability There is a direct relation between the amount of jitter in the system and the number of possible execution scenarios. Many execution scenarios require larger testing e orts. Thus, the larger the jitter is, the harder it is to test the system ....

Henrik Thane and Hans Hansson. Using Deterministic Replay for Debugging of Distributed RealTime Systems. In the 12th Euromicro Conference on Real-Time Systems, pages 265 - 272. IEEE Computer Society, June 2000.

....facility is accessible. 4.5.2 Software Monitoring Similarly to the cyclic debugging approach described above, software implemented monitoring is also vulnerable to the probe e#ect. That probe e#ect may, however, be avoided by allowing traces to remain inside the release version of the program [60] [62, s3.1;p51] Remaining probes will of course cause performance degradation, but one may argue that they shall remain also because this allows us to introduce a form of black box to the software, similar to that of airplanes. The black box may then be used if a released program experience a ....

....otherwise it is not, and must be recreated during the reproduction. The incremental replay approach also allows a replay session to start at a point which is not the starting point of the system. A feature which is very useful when the reference execution was long. Later, also Thane and Hansson [60] has provided this feature (see Section 5.4.3) Netzer presented a method based on the Instant Replay method in two articles published in 1993, Optimal Tracing and Replay for Debugging SharedMemory Parallel Programs and Trace Size vs Parallelism in Trace and Replay Debugging of Shared Memory ....

[Article contains additional citation context not shown here]

Henrik Thane and Hans Hansson. Using Deterministic Replay for Debugging of Distributed Real-Time Systems. In the 12th Euromicro Conference on Real-Time Systems, pages 265 -- 272. IEEE Computer Society, June 2000.

No context found.

H. Thane and H. Hansson. Using Deterministic Replay for Debugging of Distributed Real-Time Systems. In Proceedings of the 12th EUROMICRO Conference on Real-Time Systems, pages 265 -- 272. IEEE Computer Society, June 2000. XXX12 DANIEL SUNDMARK ET AL.

.... to account for race conditions and potential nondeterministic re executions, as well as non deterministic inputs [14] Earlier work on debugging of real time software has established the use of deterministic replay, a record replay solution, as a viable solution to create repeatable executions [1, 6, 13, 17, 15, 19]. When combining the deterministic replay approach with infinite loop recorders (analogous to black box recorders in airplanes) for postmortem debugging of embedded systems, it is essential that the recordings are sufficiently long and detailed in order to be able to re execute the system. ....

.... interrupts or mutual exclusion operations [1, 13, 20] For a more elaborate discussion on related work see [4] Earlier versions of our deterministic replay technique, which supported replay of interrupts, preemption of tasks and distributed transactions, have been presented previously [14, 15, 17]. However, none of those papers elaborated on how to identify starting points. 7. Conclusions In this paper, we presented a method for initiating a replay execution based on a previous reference execution. The replay execution can, deterministically, be cyclically repeated, it is possible to ....

H. Thane and H. Hansson. Using deterministic replay for debugging of distributed real-time systems. In Proceedings of the 12th EUROMICRO Conference on Real-Time Systems, pages 265 -- 272. IEEE Computer Society, June 2000.

.... System Debugging using Time Machines and Deterministic Replay In our previous work on debugging embedded real time systems, we have proposed a replay method for recording important events and data during a reference execution and to reproduce the execution in a subsequent replay execution [15][14] 7] As for most replay debugging methods, if the reference execution fails, the replay execution can be used to reproduce the failure repeatedly in a debugger environment in order to track down the bug. Our method allows replay of real time applications nmning on top of standard commercial ....

Thane H and Hansson H. Using Deterministic Replay for Debugging of Distributed Real-Time Systems. EUROMICRO Conference on Real-Time Systems, Pp 265272, June 2000.

....that the stacks would be large enough to store the state in addition to the already required stack size, given by the user. It will also be more difficult to monitor the system if a stack is used. Since the Asterix kernel is designed to support monitoring by using determin istic replay (DR) [17], the state needs to be stored where it is easily accessible. For each task a memory area is assigned for storage of the state. This memory area is generally called Task Contrvl Block (TCB) The first approach was to create a data structure that consisted of the necessary fields, and use this ....

....so called critical sections. These sections requires that the system is complete and that the resources involved in the section remains unchanged after the measurement, otherwise the section must be measured again. 4.3. 8 Jitter Execution time jitter is an unwanted issue in a real time system [17]. In order to minimized the jitter, sev eral design choices have been included in Asterix. Jitter minimization regarding kernel overhead is achieved by letting all paths in the kernel code have, as close as it can be, the same execution time. This is done by implementing dummy paths and dummy ....

H. Thane and H. Hansson. Using deterministic replay for debugging of distributed real-time systems. In proceedings of the 12th Euromicro Conference on real-time systems , Stockholm, June 000., 2000.

No context found.

Thane H. and Hansson H. Using Deterministic Replay for Debugging of Distributed RealTime Systems. In proceedings of the 12 Euromicro Conference on Real-Time Systems (ECRTS'00), Stockholm, June 2000.

....are arbitrary) The complexity and the testability of a system is thus strictly dependent on the infrastructure and the assumed failure semantics. That is, with our fault hypothesis the number of execution orderings would approach infinity if tasks with arbitrary arrival times were allowed. In [25][22] we elaborate on this issue and present a technique using on line recording and offline replay of task interleaving that deterministically can reproduce interleaving failures. 6 CONCLUSIONS We have in this paper introduced a novel method for deterministic testing of multitasking and distributed ....

Thane H. and Hansson H. Using Deterministic Replay for Debugging of Distributed Real-Time Systems. In Euromicro Conference on Real-Time Systems (ECRTS'00), Stockholm, June 2000.

....to jobs, and or due to synchronization errors, i.e. jobs can only interfere via specified interactions. This means that interleaving failures, e.g. memory corruptions are not considered. To handle interleaving failures other techniques need to be applied, e.g. deterministic replay debugging [15]. 3 EXECUTION ORDER ANALYSIS In this section we present a technique for identifying all the possible orders of execution for sets of jobs conforming to the task model of section 2. We first begin with a definition of execution orderings, then continue with a definition of the execution order ....

.... has an exponential relation to the jitter, as identified by Thane and Hansson [14] 13] For those that are interested we have tools for deriving the set of execution orderings as well as a real time operating system, Asterix [t7] which has the necessary infrastructure for monitoring, debugging [15] and testing of real time systems as described in this paper. Future research would be to investigate how to resolve execution order analysis with online synchronization protocols, other than PCEP, and to investigate how the testability of those protocols relate to PCEP, and offsets. Another ....

Thane H. and Hansson H. Using Deterministic Replay for Debugging of Distributed Real-Time Systems. In proceedings of the 12th Euromicro Conference on Real-Time Systems (ECRTS'00), Stockholm, June 2000.

....the kernel considered memory consumption and minimized the kernel and the application memory footprints. # Monitoring support. Built in monitoring support makes it possible to use of state of the art testing, and debugging tools like deterministic testing [14] and deterministic replay debugging [15] and visualization. The kernel also provides facilities for measuring execution times of tasks with a high resolution. # Wait and Lock free communication. The kernel supports a communication type, using buffers, that decreases the need for explicit synchronization using e.g. semaphores. As a ....

....replay debugging and deterministic testing of the target system. 3. 1 Deterministic replay Deterministic replay is a software based technique proprietary to the Asterix development environment that allows reproducible debugging of single tasking, multitasking, and distributed real time systems [15]. During runtime, information is recorded with respect to interrupts, task switches, timing, and data. The system behavior can then be deterministically reproduced off line using the recorded information. A standard debugger can be used without the risk of introducing temporal side effects, and we ....

Thane H. and Hansson H. Using Deterministic Replay for Debugging of Distributed Real-Time Systems. In proceedings of the 12 th Euromicro Conference on Real-Time Systems (ECRTS'00), Stockholm, June 2000.

No context found.

Thane H. and Hansson H. Using Deterministic Replay for Debugging of Distributed RealTime Systems. In proceedings of the 12 th Euromicro Conference on Real-Time Systems (ECRTS'00), Stockholm, June 2000.

No context found.

Henrik Thane and Hans Hansson. Using deterministic replay for debugging of distributed real-time system. In Proceedings of the 12th euromicro conference on real-time systems, pages 256--272, Stockholm, June 2000. IEEE Computer Society, IEEE Computer Society Press.

No context found.

H. Thane and H. Hansson. Using Deterministic Replay for Debugging of Distributed Real-Time Systems. In Proceedings of the 12th EUROMICRO Conference on Real-Time Systems, pages 265 -- 272. IEEE Computer Society, June 2000.

No context found.

H. Thane and H. Hansson. Using Deterministic Replay for Debugging of Distributed Real-Time Systems. In Proceedings of the 12th EUROMICRO Conference on Real-Time Systems, pages 265 -- 272. IEEE Computer Society, June 2000.

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