发布时间：2025-04-30

论文名称：PWLM3-based Automatic Performance Model Estimation Method for HDFS Write and Read Operations

发表刊物：FUTURE GENERATION COMPUTER SYSTEMS

摘要：There is a growing need for the development of an automatic performance model estimation method for Hadoop Distributed File System (HDFS) write and read (W/R) operations to cope with constant software improvement and updates, variations in parameters settings, hardware heterogeneity and their Quality of Service (QoS) evaluation. Current research based on linear system modeling methodology has a limited ability to describe the nonlinear characteristics of HDFS performance, which is an obstacle in achieving effective performance and even QoS estimation. In order to deal with this challenge, a piecewise-linear multi-model modeling (PWLM3)-based automatic performance model estimation method is proposed for HDFS W/R operations. Specially, several definitions and two strategies are presented to address four key issues when introducing PWLM3 into HDFS performance modeling. Experimental results demonstrate that the proposed method provides a good understanding and description of HDFS nonlinear characteristics and achieves better identification precision than adopting a single linear system model.

合写作者：田锋、马天、董博、郑庆华

是否译文：否

发表时间：2015-01-19