Performance Evaluation of Apache Storm With Writing Scripts

Abstract

With the exponential growth of stream data emanating from a variety of sources, today, big data presents a new era in data exploration and usage. The understanding of the performance of real-time stream data processing technologies has become a key pre-requisite while considering any deployment. Although many technologies for stream data analytics exist, little is known about the impact of writing scripts on their performance. Using a MapReduce programming model on a pseudo cluster, we conduct a word count, experimental evaluation of Apache Storm using five writing scripts; English, Arabic, Hindi, Chinese and Japanese. We define our word count as the number of time a word is written in a body of text. Static and structured data made up of 300 English sentences is translated into the scripts under study and loaded into Apache Storm. The results show that Apache Storm analyzes, English, Arabic and Hindi script sentences with ease as compared to those written in Chinese and Japanese script. Apache Storm also executes and distinguishes individual words and performs a word count on English, Arabic and Hindi sentences faster than those written in Chinese and Japanese. However, in terms of processing, the reverse is true, sentences written in Chinese and Japanese are processed faster than those written in English, Arabic and Hindi scripts.