Implementasi Hadoop: Studi Kasus Pengolahan Data Peminjaman Perpustakaan Universitas Kristen Petra

Abstract

than SQL is the general idea of this Hadoop implementation. The advancement of technology generates growing amount of data and demands a new method to process the big data. The performance of this hadoop implementation was also compared with that of SQL to prove hadoop’s novelty in processing big data. Moreover different hadoop’s implementations – such as various number of nodes, use of a combiner, and use of different block sizes – were evaluated.

Hadoop was implemented for five queries (or problems) in processing the library circulation data. Those five problems are finding the numbers of borrowing transactions categorized by the audio-video types, collection types, titles, locations, and users’ departments.

Some conclusions can be drawn based on the hadoop mapreduce implementation. Hadoop’s performance tops SQL’s when large data are being processed. The more the number of computer nodes, the faster the mapreduce application is to complete its execution. Use of a combiner can speed up the application’s execution. The arrangement with full data blocks can give better execution time than that with non-full data blocks does. In this hadoop implementation, the execution time using the block size of 128 MB is smaller than that of 28 MB and 512 MB.