Open AccessScheduling the cluster server node shutdown based on the hierarchical and k-means clustering combinations
Author(s) -
Ginanjar Aji Sudarsono,
Eddy Prasetyo Nugroho,
Rizky Rachman Judhie Putra
Publication year - 2019
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1280/3/032039
Subject(s) - computer science , cluster analysis , scheduling (production processes) , computer network , server farm , real time computing , node (physics) , duration (music) , cluster (spacecraft) , operating system , server , client–server model , engineering , mathematics , artificial intelligence , mathematical optimization , art , literature , structural engineering
The problem of server clusters is the use of electrical power. The data center in Indonesia consumed 1.5% of the national generating capacity in 2014. To solve this problem is to turn off the server node on the server cluster. With the turn off server nodes on server clusters scheduling, it’s expected to determine the time and duration of turning off server nodes. When the server cluster is running, data retrieval takes one, five and 15 minutes average load which is done every minute. Then the collected data is clustered using a combination of average linkage hierarchical clustering and K-Means clustering. The results of this clustering produce three clusters load averages that are “low”, “medium” and “high”. Load averages that included into the “low” category are sorted by the time of data retrieval to get the time and duration to turn off at each node. The results of the research result in scheduling turning off node one is 14.09 - 14.28, at node two is 14.47 - 06.57 and at node three is 06.57 - 07.18. Turn off server node scheduling is reduces the use of electrical power from 2,528 kWh to 2,519 kWh and doesn’t affect server quality parameters.