Hotspot diagnosis on logical level

Abstract

Hotspots in data center have been attributed to an increase in equipment failures, which causes system down time and business loss. In maintenance of IT equipment, removing hotspot with minimal cooling power cost is both ecological and financial benefits. Thus, identifying root cause of hotspots in data center is essential. A variety of factors can cause the issue, and therefore, diagnosing root cause of hotspot for removing and preventing it becomes complicated. This paper proposes a technique which supports diagnose the hotspot from logical level relative to physical conditions. In the proposed technique, performance and configuration data of each potential factor are analyzed with hotspot trochoid automatically. Effect and impact of the factor are synthesis evaluated for each component to diagnose whether it is related to the hotspot emerging or not. The diagnosis results are collected for suspects of creating hotspot in the target ambient, and then, the suspects in the whole system can be ranked for helping to identify the root cause. Hotspot diagnosis from the logical level is also useful to improve the hotspot diagnosis on other external physical conditions factors, when the logical causes are excluded by the proposed approach. © 2011 IFIP.