A flexible architecture integrating monitoring and analytics for managing large-scale data centers

摘要: To effectively manage large-scale data centers and utility clouds, operators must understand current system application behaviors. This requires continuous, real-time monitoring along with on-line analysis of the captured by system, i.e., integrated analytics -- Monalytics [28]. A key challenge such integration is to balance costs incurred associated delays, against benefits attained from identifying reacting to, in a timely fashion, undesirable or non-performing states. paper presents novel, flexible architecture for which trade-offs are easily made dynamically constructing software overlays called Distributed Computation Graphs (DCGs) implement desired functions. The prototype implementing this evaluated motivating use cases small scale center experiments, series analytical models used above at large scales. Results show that approach provides flexibility needed meet demands autonomic management considerably better performance/cost than traditional brute force solutions.