Rapid Deployment for Machine Learning in Educational Cloud

Abstract

In the cloud era, the acquisition of new cloud skills is a constant requirement of IT specialists. Educational organizations such as universities have a need to provide educational cloud curriculums for their students. In our current research, we are constructing a private cloud based on super-saturation, which is defined as the allocation of a much greater amount of logical resources than physical resources. Super-saturated clouds therefore realize up to 10 times more running instances than conventional clouds. While the performance of super-saturated clouds decreases somewhat compared with conventional clouds, their costs also greatly decrease. Moreover, in the post-cloud era, i.e., the big data era, data scientists will be increasingly required to process big data in the cloud. Mahout and Hadoop are two popular tools used in the fields of data science and machine learning. However, a certain level of skill is required to build such machine learning systems, and because it takes a long time to build such systems, the curriculums available to learners are limited. In this paper, we propose a method of rapid deployment for machine learning systems in the educational cloud. We show that our proposed method can reduce the required preparation time.