BigHOST: Automatic Grading System for Big Data Assignments

Abstract

With the increasing popularity of online courses in Big Data, Data Science, and Machine Learning, the need for an efficient and reliable grading solution for assignments has become evident. Existing solutions for auto-grading assignments are limited to simple coding assignments and are unable to handle the complexity, variety, and volume of data required in Big Data applications. In order to address this need, we propose BigHOST, a custom-designed auto-grader for Big Data assignments. BigHOST employs a simple yet vertically scalable, fault-tolerant and parallel processing architecture, making it efficient and reliable for grading big data assignments. Optimizations in the architecture further result in lower execution time per submission and reduced cost of hosting on cloud platforms. Experimental results and scalability analysis demonstrate the effectiveness of the proposed architecture, with BigHOST achieving more than five times the throughput in processing big data submissions.