Revisiting Deterministic Multithreading Strategies

Domaschka, Jörg; Schmied, A.I.; Reiser, Hans P.; Hauck, Franz J.

doi:10.1109/ipdps.2007.370415

“…However, in ongoing work [19] we demonstrate that code analysis and transformation allows improving concurrency on the basis of prediction of future synchronization steps.…”

Section: Criteriamentioning

confidence: 94%

“…The poor performance of MAT can be alleviated by the introduction of yield operations, which enable a selection of a new primary thread without reaching an implicit scheduling point. Another approach for optimizing MAT is the use of source-code analysis to predict synchronization behaviour [19].…”

Section: Local Computationsmentioning

confidence: 99%

Multithreading Strategies for Replicated Objects

Domaschka¹,

Bestfleisch²,

Hauck³

et al. 2008

Middleware 2008

Self Cite

3

0

View full text Add to dashboard Cite

Abstract. Replicating objects usually requires deterministic behaviour for maintaining a consistent state. Multithreading is a critical source of non-determinism, completely unsupported in most fault-tolerant middleware systems. Recent publications have defined deterministic scheduling algorithms that operate at the middleware level and allow multithreading for replicated objects. This approach avoids deadlocks, improves performance, and makes the development better resemble that of non-replicated objects. This paper surveys those algorithms and analyses their differences. It also defines extensions to two efficient multithreading algorithms to support nested invocations and condition variables with time-bounded wait operations similar to the Java synchronisation model. In addition, we provide an experimental evaluation and performance comparison of the algorithms, indicating the areas in which each algorithm performs best. We conclude that replication middleware should implement reconfigurable multithreading strategies, as there is no optimal one-size-fits-all solution.

show abstract

“…Multithreaded replica systems add fault tolerance by executing many replicas of a program so that nodes may fail without interruption in service (Basile et al 2002;Domaschka et al , 2007Saha and Dutta 1993;Karl et al 1998;. This type of technique relies on determinism so that replicas remain synchronized.…”

Section: Related Workmentioning

confidence: 99%

Kendo

Olszewski

¹

,

Ansel

²

,

Amarasinghe

³

2009

View full text Add to dashboard Cite

Although chip-multiprocessors have become the industry standard, developing parallel applications that target them remains a daunting task. Non-determinism, inherent in threaded applications, causes significant challenges for parallel programmers by hindering their ability to create parallel applications with repeatable results. As a consequence, parallel applications are significantly harder to debug, test, and maintain than sequential programs. This paper introduces Kendo: a new software-only system that provides deterministic multithreading of parallel applications. Kendo enforces a deterministic interleaving of lock acquisitions and specially declared non-protected reads through a novel dynamically load-balanced deterministic scheduling algorithm. The algorithm tracks the progress of each thread using performance counters to construct a deterministic logical time that is used to compute an interleaving of shared data accesses that is both deterministic and provides good load balancing. Kendo can run on today's commodity hardware while incurring only a modest performance cost. Experimental results on the SPLASH-2 applications yield a geometric mean overhead of only 16% when running on 4 processors. This low overhead makes it possible to benefit from Kendo even after an application is deployed. Programmers can start using Kendo today to program parallel applications that are easier to develop, debug, and test.

show abstract

Kendo

Olszewski

¹

,

Ansel

²

,

Amarasinghe

³

2009

Proceedings of the 14th International Conference on Architectural Support for Programming Languages and Operating Systems

View full text Add to dashboard Cite

Although chip-multiprocessors have become the industry standard, developing parallel applications that target them remains a daunting task. Non-determinism, inherent in threaded applications, causes significant challenges for parallel programmers by hindering their ability to create parallel applications with repeatable results. As a consequence, parallel applications are significantly harder to debug, test, and maintain than sequential programs. This paper introduces Kendo: a new software-only system that provides deterministic multithreading of parallel applications. Kendo enforces a deterministic interleaving of lock acquisitions and specially declared non-protected reads through a novel dynamically load-balanced deterministic scheduling algorithm. The algorithm tracks the progress of each thread using performance counters to construct a deterministic logical time that is used to compute an interleaving of shared data accesses that is both deterministic and provides good load balancing. Kendo can run on today's commodity hardware while incurring only a modest performance cost. Experimental results on the SPLASH-2 applications yield a geometric mean overhead of only 16% when running on 4 processors. This low overhead makes it possible to benefit from Kendo even after an application is deployed. Programmers can start using Kendo today to program parallel applications that are easier to develop, debug, and test.

show abstract

Revisiting Deterministic Multithreading Strategies

Cited by 4 publications

References 9 publications

Multithreading Strategies for Replicated Objects

Multithreading Strategies for Replicated Objects

Kendo

Kendo

Contact Info

Product

Resources

About