Learning Heterogeneous Cloud Storage Configuration for Data Analytics
Conference on Systems and Machine Learning (SysML), 2018
Abstract
The public cloud, with its promise of elasticity and reduced total cost of ownership, is experiencing unprecedented growth. However, performance and cost efficiency are only achieved by choosing a suitable configuration for each given application.
For data-intensive analytics commonly hosted in the cloud, the choice of storage is essential. Selecta learns near-optimal VM and storage configurations for analytics applications and makes recommendations to satisfy user-specified performance-cost objectives.