Questa pagina descrive il corso di dottorato “Model-free and Model-based Auto-Scaling Techniques for Distributed Applications“, tenuto dal Prof. Gabriele Russo Russo
Date
- 5 giugno 14:00-16:00
- 12 giugno 14:00-16:00
- 14 giugno 14:00-16:00
- 22 giugno, 14:00-18:00
Abstract
Unpredictable workload variations as well as unstable computing and network infrastructure conditions represent major challenges for distributed applications subject to Quality-of-Service requirements. As manual (re)configuration of applications is not feasible for large-scale systems and infrastructures, applications need self-adaptation abilities to cope with changing working conditions.
In these lectures we especially focus on auto-scaling, which allows applications to dynamically acquire and release resources (e.g., CPU shares, memory or whole computing nodes) as needed. We will discuss the challenges associated with the definition of auto-scaling policies and present different approaches ranging from traditional model-based techniques (e.g., based on queueing theory) to model-free ones based on reinforcement learning, as well as hybrid approaches.