2016 International Conference on Cloud and Autonomic Computing (ICCAC) (2016)
Sept. 12, 2016 to Sept. 16, 2016
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/ICCAC.2016.12
Today, Apache Cassandra, an highly scalable and available NoSql datastore, is largely used by enterprises of each size and for application areas that range from entertainment to big data analytics. Managed Cassandra service providers are emerging to hide the complexity of the installation, fine tuning and operation of Cassandra datacenters. As for all complex services, human assisted management of a multi-tenant cassandra datacenter is unrealistic. Rather, there is a growing demand for autonomic management solutions. In this paper, we present an optimal energy-aware adaptation model for managed Cassandra datacenters that modify the system configuration orchestrating three different actions: horizontal scaling, vertical scaling and energy aware placement. The model is built from a real case based on real application data from Ericsson AB. We compare the performance of the optimal adaptation with two heuristics that avoid system perturbations due to re-configuration actions triggered by subscription of new tenants and/or changes in the SLA. One of the heuristic is local optimisation and the second is a best fit decreasing algorithm selected as reference point because representative of a wide range of research and practical solutions. The main finding is that heuristic's performance depends on the scenario and workload and no one dominates in all the cases. Besides, in high load scenarios, the suboptimal system configuration obtained with an heuristic adaptation policy introduce a penalty in electric energy consumption in the range [+25%, +50%] if compared with the energy consumed by an optimal system configuration.
Adaptation models, Optimization, Cloud computing, Throughput, Tuning, Scalability, Mathematical model
E. Casalicchio, L. Lundberg and S. Shirinbab, "Energy-Aware Adaptation in Managed Cassandra Datacenters," 2016 International Conference on Cloud and Autonomic Computing (ICCAC), Augsburg, Germany, 2016, pp. 60-71.