Reconfigurable, data-driven resource allocation in complex systems: practice and theoretical foundations

We focus on the development of a data-driven performance engineering framework to automate the process of robust, workload-aware resource allocation and management in today´s complex Internet servers. Our focus is on the development of better understanding of the workload resource demands and on the development and implementation of efficient methodologies for bottleneck identification and resource allocation at the system level. Here, we give an overview of a testbed for conducting a detailed workload characterization in multi-tiered web servers that serve dynamic pages. We present some preliminary workload characterization results that can help in identifying different resource bottlenecks and the workload conditions under which these bottlenecks are triggered. We also present preliminary results on new analytic models that one can use to model multi-tiered systems.