|BRONDOLIN ROLANDO||Cycle: XXXII |
Section: Computer Science and Engineering
Tutor: DANIEL FLORIAN
Advisor: SANTAMBROGIO MARCO DOMENICO Major Research topic
:On the management of power and performance trade-offs in distributed cloud-native infrastructuresAbstract:
In the last few years, the cloud computing model shifted the software development paradigms from a pure monolithic approach based on Virtual Machines (VMs) to more scalable approaches leveraging several open source technologies. Docker containers and container orchestrators like Kubernetes, Mesos and Swarm; streaming systems like Apache Flink, Apache Storm and Apache Spark Streaming are now at the heart of modern cloud computing applications and infrastructures. All those tools typically handle applications and services like search engines, social networks, instant payment services, webmail services, automatic translation, and more in general software as a service platforms. Given the criticality of such applications from a business point of view, they need to be extremely responsive even in case of massive amount of requests. As such, servers are provisioned to handle efficiently the peak loads of the applications at the cost of low energy efficiency at medium and low data-center load. This happens as the load of cloud services can widely vary during the day, however their power consumption has less variance due to the lack of energy proportionality of the underlying hardware. Given the scale and the latency sensitivity of cloud workloads, a systematic and accurate approach is then needed to optimally manage power consumption at scale. Energy Efficiency for Autonomic Scalable sYstems (E2ASY) aims at becoming the reference set of techniques and methodologies for the autonomic management of scalable systems towards energy-proportionality of cloud workloads. The goal is to optimize power consumption and energy consumption on the long run guaranteeing a minimum level of performance defined as a Service Level Agreement (SLA).