Fuzzy Self-Learning Controllers for Elasticity Management in Dynamic Cloud Architectures

Abstract

Abstract-Cloud controllers support the operation and quality management of dynamic cloud architectures by automatically scaling the compute resources to meet performance guarantees and minimize resource costs. Existing cloud controllers often resort to scaling strategies that are codified as a set of architecture adaptation rules. However, for a cloud provider, deployed application architectures are black-boxes, making it difficult at design time to define optimal or pre-emptive adaptation rules. Thus, the burden of taking adaptation decisions often is delegated to the cloud application. We propose the dynamic learning of adaptation rules for deployed application architectures in the cloud. We introduce FQL4KE, a self-learning fuzzy controller that learns and modifies fuzzy rules at runtime. The benefit is that we do not have to rely solely on precise design-time knowledge, which may be difficult to acquire. FQL4KE empowers users to configure cloud controllers by simply adjusting weights representing priorities for architecture quality instead of defining complex rules. FQL4KE has been experimentally validated using the cloud application framework ElasticBench in Azure and OpenStack. The experimental results demonstrate that FQL4KE outperforms both a fuzzy controller without learning and the native Azure auto-scaling.

A control theoretical view of cloud elasticity: taxonomy, survey and challenges

“…Lastly, they are also criticized more often for their associated unwanted behaviour, termed as bumpy transition, that could lead the system to an oscillatory state [27,104,105]. On the other hand, knowledge-based control solutions utilizing machine learning [67,106,107] or neural networks [99,108] provide high levels of flexibility and adaptivity. However, such flexibility and adaptivity come at the cost of long training delays, poor scalability, slower convergence rate, and the impossibility of deriving stability proof [7,18,103,109].…”

“…The authors of [47,67,97] proposed horizontal elasticity solution for Internet based systems using fuzzy systems. They all have used performance based metrics to make scaling decisions.…”

“…Furthermore, the output (number of VMs) of their approach are a predefined fixed range. Their approach is further extended in [67], where they have used fuzzy Q-Learning to learn the best elastic policies (fuzzy rules) at runtime to cope with the issue mentioned above.…”

“…This approach however, assume that all the servers are homogeneous. Furthermore, in contrast to [47,67], they also consider resource utilisation in decision making in accompanying to their performance measurements.…”

Adaptive Image Thresholding of Yellow Peppers for a Harvesting Robot

“…Moreover, it is shown that it is beneficial to use RL for classification of objects [18,21]. Reinforcement learning has also been used in many other applications, such as web applications [22,23], control and management problems [24,25], and also in grid and cloud computing [26,27].…”