The Hadoop distributed filesystem: Balancing portability and performance

TLDR

Hadoop, an open‑source MapReduce framework, employs the Java‑based HDFS to provide a portable distributed filesystem across heterogeneous hardware and software platforms. This study examines HDFS performance, identifies bottlenecks, and evaluates the trade‑offs between portability and efficiency. The authors find that scheduling delays create architectural bottlenecks, Java’s portability limits exploitation of native platform features, and HDFS’s assumptions about native storage management lead to inefficiencies across diverse systems.

Abstract

Hadoop is a popular open-source implementation of MapReduce for the analysis of large datasets. To manage storage resources across the cluster, Hadoop uses a distributed user-level filesystem. This filesystem - HDFS - is written in Java and designed for portability across heterogeneous hardware and software platforms. This paper analyzes the performance of HDFS and uncovers several performance issues. First, architectural bottlenecks exist in the Hadoop implementation that result in inefficient HDFS usage due to delays in scheduling new MapReduce tasks. Second, portability limitations prevent the Java implementation from exploiting features of the native platform. Third, HDFS implicitly makes portability assumptions about how the native platform manages storage resources, even though native filesystems and I/O schedulers vary widely in design and behavior. This paper investigates the root causes of these performance bottlenecks in order to evaluate tradeoffs between portability and performance in the Hadoop distributed filesystem.