Chunlin Wang
ABSTRACT
The cloud service providers manage very large data centers all over the world. When an abnormality occurs, various types of alarm information are triggered. The operation engineers need to quickly discover and locate all the abnormalities. To solve the problems of computing-intensive, non-real-time, and inaccurate abnormal detection and location algorithm, we propose to improve the Monte Carlo Tree Search (MCTS) based on the greedy algorithm by: 1) improving the selection of the next node in MCTS by using greedy algorithm and searching the best node with depth-first method; 2) adopting the sparse matrix to store the record of the 5-dimensional log files, then employing the subscript file to record the subscript of the 5-dimensional array, and using the subscript to access the sparse matrix to save memory space and searching time; 3) reducing calculations by pruning some branches based on the observation that the optimal node combination of present layer must belong to the search space of the best layer combination of the previous layer. The experimental results show that GMCTS algorithm reduces 40% computation time than the HotSpot in 5D data, and the correct positioning efficiency is up to 96.1%.
- Yongqian Sun, Youjian Zhao, Ya su, , “HotSpot:Anomaly Localization for Additive KPIs withMulti-Dimensional Attributes”, IEEE Access, 2018.Google Scholar
- Chen M Y, Kiciman E, Fratkin E, Pinpoint: Problem determination in large, dynamic internet services[C]//Proceedings International Conference on Dependable Systems and Networks. IEEE, 2002: 595-604.Google Scholar
- Q. Lin, J. Lou, H. Zhang, and D. Zhang, “idice: problem identification for emerging issues,” ICSE, 2016, ACM,,pp. 214–224.Google ScholarDigital Library
- Kompella R R, Yates J, Greenberg A, Fault localization via risk modeling[J]. IEEE Transactions on Dependable and Secure Computing, 2009, 7(4): 396-409.Google ScholarDigital Library
- Kandula S, Mahajan R, Verkaik P, Detailed diagnosis in enterprise networks[J]. ACM SIGCOMM Computer Communication Review, 2009, 39(4): 243-254.Google ScholarDigital Library
- Zhu Y, Kang N, Cao J, Packet-level telemetry in large datacenter networks[C]//ACM SIGCOMM Computer Communication Review. ACM, 2015, 45(4): 479-491.Google Scholar
- R. Bhagwan, R. Kumar, and R. o. Ramjee, “Adtributor: Revenue debugging in advertising systems,” NSDI, 2014, pp. 43–55.Google Scholar
- Gelly S, Kocsis L, Schoenauer M, The grand challenge of computer Go: Monte Carlo tree search and extensions[J]. Communications of the ACM, 2012, 55(3): 106-113.Google ScholarDigital Library
- Silver D, Schrittwieser J, Simonyan K, Mastering the game of go without human knowledge[J]. Nature, 2017, 550(7676): 354.Google ScholarCross Ref
- Bahl P, Chandra R, Greenberg A, Towards highly reliable enterprise network services via inference of multi-level dependencies[J]. ACM SIGCOMM Computer Communication Review, 2007, 37(4): 13-24.Google ScholarDigital Library
- Zhihua Zhou. Machine Learning[M]. Tsinghua University Press, 2016:30-38.Google Scholar
- Zhuang, Y., Gessiou, E., Portzer, S., Fund, F., Muhammad, M., Beschastnikh, I., & Cappos, J. (2014). Netcheck: Network diagnoses from blackbox traces. In 11th {USENIX} Symposium on Networked Systems Design and Implementation ({NSDI} 14) (pp. 115-128).Google Scholar
Recommendations
Monte Carlo Tree Search with Metaheuristics
Artificial Intelligence and Soft ComputingAbstractMonte Carlo Tree Search/Upper Confidence bounds applied to Trees (MCTS/UCT) is a popular and powerful search technique applicable to many domains, most frequently to searching game trees. Even though the algorithm has been widely researched, there ...
Population based Monte Carlo tree search hyper-heuristic for combinatorial optimization problems
Hyper-heuristics aim to automate the heuristic selection process in order to operate well across different problem instances, or even across different problem domains. A traditional hyper-heuristic framework has two levels, a high level strategy and a ...
Multiagent Monte Carlo Tree Search
AAMAS '19: Proceedings of the 18th International Conference on Autonomous Agents and MultiAgent SystemsMonte Carlo Tree Search (MCTS) is a best-first search which is efficient in large search spaces and is effective at balancing exploration versus exploitation. In this work, we introduce a novel extension for MCTS, called Multiagent Monte Carlo Tree ...
Comments