A Performance Auto-Tuner for Cloud Systems called ClassyTune

PROJECT TITLE :

ClassyTune: A Performance Auto-Tuner for Systems in the Cloud

ABSTRACT:

Tuning the system's performance can make it run more efficiently, which in turn enables a reduction in the amount of Cloud Computing resources that are required to support an application. Automating performance tuning for complicated cloud-based systems is necessary because the number of parameters and complexity of systems continues to grow at an alarming rate. Methods that are considered to be state-of-the-art use either an approach that is data-driven or one that is driven by experience when it comes to tuning. Since it can be applied to a greater variety of contexts, data-driven tuning is gaining an increasing amount of attention. However, the data-driven methods that are currently available are unable to adequately address both the limited sample size and the high dimensionality issues simultaneously. ClassyTune is a data-driven automatic configuration tuning tool for cloud systems, and we are pleased to present it. ClassyTune uses a model of classification that is based on Machine Learning in order to perform autotuning. Because of this exploitation, it is possible to induce more training samples without simultaneously increasing the input dimension. Experiments conducted on seven widely used systems hosted in the cloud have demonstrated that ClassyTune is superior to both expert tuning and the most cutting-edge auto-tuning solutions thanks to its ability to effectively tune system performance to be seven times higher for high-dimensional configuration spaces. In addition, we describe a use case that illustrates how performance tuning can enable a reduction of 33 percent in the amount of computing resources required to run an online service that is stateless.

Did you like this research project?

To get this research project Guidelines, Training and Code... Click Here