Self-organizing Maps versus Growing Neural Gas in Detecting Anomalies in Data Centres

Abstract

Reliability is one of the key performance factors in Data Centers. The out-of-scale energy costs of these facilities lead Data Center operators to increase the ambient temperature of the data room to decrease cooling costs. However, increasing ambient temperature reduces the safety margins and can result in a higher number of anomalous events. Anomalies in the Data Center need to be detected as soon as possible to optimize cooling efficiency and mitigate the harmful effects over servers. This paper proposes the usage of clustering-based outlier detection techniques coupled with a Trust and Reputation System (TRS) engine to detect anomalies in Data Centers. We show how SOM and GNG can be applied to detect cooling and workload anomalies respectively in a real Data Center scenario with very good detection and isolation rates, in a way that is robust to the malfunction of the sensors that gather server and environmental information.

Thermal Management in Large Data Centres: Security Threats and Mitigation

“…The method, which was notably based on thermal cameras and traditional temperature monitoring devices, demonstrated sufficient accuracy in anomalous cases. Later, and responding to the growing need for fast online detection of fault cases, [3,41] introduced the use of Self-Organizing Maps and reputation systems.…”

The GNG neural network in analyzing consumer behaviour patterns: empirical research on a purchasing behaviour processes realized by the elderly consumers

“…The GNG network finds application in various fields of science today: i.e. medical diagnostics (Ogura et al 2003;Netto et al 2012), analysis of shopping habits (Decker and Monien 2003;Decker 2005; Migdał-Najman and Najman 2013), pattern recognition (Datta et al 2001;Frezza-Buet 2008;Sun and Yang 2010;Viejo et al 2012), the device control process (Pucci 2010;Zapater et al 2015), and social sciences (Memmert and Perl 2009). In recent years, the research on the development of GNG networks was focused on algorithm optimization (Chávez et al 2011;Aljobouri et al 2018) application in large data systems (parallel algorithms) (Vojáček and Dvorský 2013;Fliege and Benn 2016), model construction algorithms (supervised GNG) (Duque-Belfort et al 2017).…”

MAGNETIC: Multi-Agent Machine Learning-Based Approach for Energy Efficient Dynamic Consolidation in Data Centers

“…H IGH energy consumption and performance variability problems are two major challenges in modern cloud data centers, as they greatly affect operational expenses, total cost of ownership and revenue [1]. Global data center electricity usage accounted for 1.1-1.5% of total electricity use in 2010 [2] and increases at yearly rate of 2.1% [3].…”

“…Application performance variability is one of the major challenges in data centers, and may originate from both internal and external sources, e.g., aged or failing hardware, thermal control, orphan processes or operating system issues [5]. Most of these anomalies lead to performance degradation during host operation [1] and, therefore, to QoS degradation. Detecting performance variability is usually accomplished by means of realtime continuous monitoring and analyzing system logs.…”