2026-09-21T03:19:24Z https://www.ideals.illinois.edu/oai-pmh

oai:www.ideals.illinois.edu:2142/20480 2023-07-10 col_2142_5131 col_2142_10761 com_2142_5130 com_2142_10755 com_2142_234

Belford, Geneva G. Boeheim, Chidori Kawamura 2011-05-07T12:40:26Z 2011-05-07T12:40:26Z 10000-01-01 1991 Distributed deadlock is a state where there exists among some processes running on different computers a cyclic wait to acquire some resources such that no process can proceed. Since a deadlock is a state that persists unless it is solved by some method, and the number and size of deadlocks increase substantially as the concurrency level or number of sites in a distributed system grows, we need an efficient approach to solve this problem. Distributed deadlock resolution consists of choosing a victim, aborting and restarting it. There has been no detailed study of resolution, despite the fact that deadlock detection does not complete its task without resolution. The performance analysis of different resolution strategies that have been proposed and new approaches for distributed deadlock resolution are the subject of this study. Different heuristic strategies (rules) to choose victims to break deadlocks are discussed and the direct and indirect goals that the rules were hypothesized to achieve are described. Various simulation runs (experiments) were conducted to analyze in depth the performance of the rules. The system throughput and the overhead of running the rules are evaluated and the effectiveness of each rule in achieving its goals are compared. The assessment of each rule under different system conditions is used to evaluate the possibility of incorporating more than one rule in deadlock resolution. This leads to the suggestion for use of a first-principles expert system to monitor and diagnose distributed systems, including the resolution of problems such as deadlocks. The rule base of such an expert system would include rules to choose the optimum victim to resolve a distributed deadlock. Made available in DSpace on 2011-05-07T12:40:26Z (GMT). No. of bitstreams: 2 license.txt: 4922 bytes, checksum: 910b249b4beec47e7ab768910c8f966f (MD5) 9136547.pdf: 6629682 bytes, checksum: 43cac80fd01e2880662c2c60b959ef38 (MD5) Previous issue date: 1991 Item marked as restricted to the 'UIUC Users [automated]' Group (id=2) by Howard Ding (hding2@illinois.edu) on 2011-05-07T14:44:10Z Item is restricted indefinitely. Restriction data tranferred 2014-07-01T11:19:25-05:00 Original Data Group with Access UIUC Users [automated] Release Date: none Reason: ETDs are only available to UIUC Users without author permission ETDs are only available to UIUC Users without author permission U of I Only AAI9136547 (UMI)AAI9136547 http://hdl.handle.net/2142/20480 eng Copyright 1991 Boeheim, Chidori Kawamura Artificial Intelligence Computer Science A performance study of deadlock resolution in distributed systems text Computer Science Computer Science University of Illinois at Urbana-Champaign Dissertation Ph.D.