Abstract :
To guarantee the vision of Quality of Service (QoS) different goals in terms of SLAs have to be dynamically met between the Cloud provider and the customer (Breskovic et al., 2013). This SLA enactment should involve little human-based interaction in order to guarantee the scalability and efficient resource utilization of the system. To achieve this we start from Autonomic Computing, examine the autonomic control loop and adapt it to govern Cloud Computing infrastructures. We propose an approach to manage Cloud infrastructures by means of Autonomic Computing. The basic structure of the autonomic systems is represented by a control loop that monitors (M) Cloud parameters, analyses (A) them, plans (P) actions and executes (E) them; the full cycle is known as MAPE. MAPE-K loop stores knowledge (K) required for decision-making in a knowledge base (KB) that is accessed by the individual phases. This talk addresses the research question of finding a suitable KM system (i.e., a technique of how stored information should be used) and determining how it interacts with the other phases for dynamically and efficiently allocating resources. We first hierarchically structure all possible adaptation actions into so-called escalation levels (Chimaobi et al., 2011). We then focus on one of these levels by analysing monitored data from virtual machines and making decisions on their resource configuration with the help of knowledge management (KM). The monitored data stems both from synthetically generated workload categorized in different workload volatility classes and from a real-world scenario: scientific workflow applications in bioinformatics. As KM techniques, we investigate two methods, Case-Based Reasoning (CBR) and a rule-based approach. We design and implement both of them and evaluate them with the help of a simulation engine. Simulation reveals the feasibility of the CBR approach and major improvements by the rule-based approach considering SLA violations, resource utilization, the number of necessary reconfigurations and time performance for both, synthetically generated and real-world data. Both approaches are evaluated with the real world data from the life science domain by using the traces of RNA sequencing. RNA sequencing represent a new challenging approach for the evaluation and profiling of the gene structures. With the rapid development of high-throughput technologies in recent years, huge amounts of data are being generated and stored in databases in the field of life science, which necessitates significant advances in computing capacity and performance. RNA sequencing has the potential to transform how gene structure and gene expression profiling are studied. Scientific workflow applications are crucial in enabling scientists to determine important information from those huge amounts of stored data. Existing workflow applications are process-or data-, rather than resource oriented. Thus, they lack efficient computational resource management capabilities, such as those provided by Cloud computing environments. Insufficient computational resources disrupt the execution of workflow applications, wasting time and money. To address this issue, advanced resource monitoring and management strategies are required to determine the resource consumption behaviors of workflow applications for a dynamical allocation and deallocation of resources (Maurer et al., 2013). Thus, we utilize the knowledge management strategies for Clouds (rule and CBR-based) to manage computational resources for workflow applications in order to guarantee their performance goals and their successful completion (Maurer et al., 2013).
