[XSL-LIST Mailing List Archive Home] [By Thread] [By Date]

[xsl] Performance Answer

Subject: [xsl] Performance Answer
From: "Ihe Onwuka ihe.onwuka@xxxxxxxxx" <xsl-list-service@xxxxxxxxxxxxxxxxxxxxxx>
Date: Fri, 5 Jun 2015 11:36:52 -0000

What was the question.

The task - Upconverting a 0.5GB text file that Saxon cannot swallow whole,
so some sort of fragmentation was necessary.

The cruder fragmentation split the file into 700 files of widely varying
sizes as it depended upon the amount of data per sub-year.

A slightly more involved fragmentation splitting the files on year only
resulted in 139 files, again widely varying in size.

So would running 700 transforms be faster than running 139, I thought not.

This was run on Ubuntu Linux, the transformations were parallelized with
xargs such that a maximum of 4 would run simultaneously.

Turns out it wasn't even close. The 700 file fragmentation ran in 20
minutes, the 139 file one took 54.

Current Thread