Two Browsers, Two Computers, Two Linuxes: Streaming Audio

07 March 2011

Despite numerous improvements in Linux, during the past decade, it consistently underperforms products by Microsoft (M$). This distinction may be blurred in newer computers, but on older machines, Linux dependably performs routine tasks more slowly than M$ software.

In some settings, however, this relative lethargy is not a significant handicap. One such application,using older computers, with Linux, where perceived diminution in responsiveness is a relatively insignificant distraction, is streaming audio reception. The following test was devised to compare various operating systems and web browsers tasked with the chore of receiving web broadcasts from radio stations around the world, employing any one of three non-proprietary formats: OGG, aac+, or mp3. The goal of the test was to measure the times needed to receive broadcasts from each of twelve different stations, found at Mike's Classical Music:
http://www.listenlive.eu/classical.html

To emulate the concept of using one's computer concurrently to perform some sort of additional task, while listening to music,a background task is invoked-->downloading a 4GB DVD from the Debian web site. This has the net effect of slowing down reception, upon changing stations, compared with operating the computer engaged in no task other than receipt of broadcast streaming audio.

Procedure: After booting, the background task is begun: downloading a 4 gigabyte file. The browser under investigation is engaged, and directed to Mike's web site above. VLC is successively invoked and then closed, by clicking on first, an mp3 icon, then an aac+ icon, and finally an ogg icon, all three icons representing web sites not included in the list of twelve sites, below. VLC is closed again, and the trial then commences. Each trial consists of twelve different times corresponding to the time needed to hear music after clicking successively on the icon representing each of these twelve sites, closing VLC after each time has been recorded. The twelve times are then summed to represent total time for that particular trial. After discarding the downloaded 4 gigabyte file, the machine is rebooted, a new 4 gigabyte download task begun,
and the next browser is tested. Each browser is tested in this fashion, three times, on three different days, and the total times are then summed for each browser on each distribution on the various computers. Each number below thus represents the sum of thirty six different measurements.
Computer specifications: Both have default video resolution of 1280 x 1024 with Win 98
Results:
Massage: here is some of the raw data normalized by incorporating the cpu and memory bandwidth:
Interpretation: Of note is the distinction between the times needed for reception under XP, a 32 bit operating system, running on C8, with a 64 bit cpu, versus times measured on the same computer using a 64 bit version of CrunchBang Linux. PCLinuxOS (32 bit) versus CrunchBang (64 bit), both running Konqueror, demonstrates the superior advantage of 64 bit operations, versus 32 bit, (else dual core compared with single core), even in this relatively unchallenging data processing environment: streaming audio reception while concurrently downloading a large file. The very same PCLinuxOS, with identical Konqueror browser, on the same, single core, 32 bit machine, outperforms by 20%, the 32 bit version of CrunchBang running the same browser, so we are relatively confident that the 16% improvement in 64 bit operations, compared with 32 bit functionality, does not reflect a superiority in CrunchBang's implementation of Linux, per se, compared with that of PCLinuxOS. The improved performance observed, in my opinion, is due not to sophistication of contemporary 64 bit Linux, but to the architectural advantages offered by the cpu. How else can one explain the observation that 32 bit XP is twice as fast as 64 bit Linux?

CAI ENG