So what does all of this mean? It means that OpenOffice, though a fine, free suite, has a long ways to go in terms of application load time to be really competitive. While the operating systems (particularly Mandriva and Windows) are basically comparable, OpenOffice was a real performance drag. 211 seconds for the stress test on the Metasys machine was especially painful. One moderately redeeming note for OpenOffice: once a single application is open, all of the others launch very quickly from within OO. For example, choosing New Spreadsheet from the File menu of the word processing application had a new spreadsheet up within 5 seconds. However, this certainly remains a sticking point. In terms of real usability, some of this comes down to personal preference and customizability. What makes sense to the customers and what role to local groups want to have in localization of the software? How much does licensing matter? And what just feels right? For me and about half of my students, Mandriva was the CMPC of choice. For the remaining half, Windows was the preferred OS. As we've seen over the last few posts, this is a religious issue. The take home message, though? These little guys work and they work really well. Intel, can we please have some here in your "established markets"?
Test Operating System (time in seconds) Windows XP Professional Mandriva Linux Metasys Linux Launch to login N/A - logins disabled on test system 69 90 Login to fully loaded (all services running; ready to use) N/A 32 22 Total time to system usable from startup 116 101 112 Launch OpenOffice Writer N/A 35 49 Launch OpenOffice Impress N/A 24 24 Launch OpenOffice Calc N/A 26 27 Launch Firefox (and load standard Firefox/Google homepage) 14 9 10 Launch Internet Explorer (and open default MSN homepage) 8 N/A N/A Launch Microsoft Word 4 N/A N/A Launch Microsoft Excel 6 N/A N/A Launch Microsoft PowerPoint 3 N/A N/A Copy 118MB file to the hard drive (OpenOffice Install file) 133 139 137 Stress test 17 103 211
I feel like I should set the screensavers to a scrolling marquee of "Don't ever change" or "You rock," standard yearbook fare for saying goodbye to carbon-based classmates. Tonight, however, I'm packaging up 3 Intel Classmate PCs (CMPC) to ship back to California, and I'm really going to miss these little computers. More significantly, so are my kids and my students. They've been handy little tools in class and at home and have shown just how realistic 1:1 computing (or at least drastically increased access to computing for school kids) can be. Before I wrap up the series on the Classmates, though, I'd like to run through some performance numbers. I have lots of seat-of-the pants impressions, but the following table actually details times to complete various tasks on the computers. Since I have three different operating systems available (Mandriva Linux, Metasys Linux, and Windows XP Professional), I've included columns for all three. In the case of the Windows machine, I installed Firefox for comparison to their Linux counterparts; There wasn't enough room on the hard drive to install OpenOffice. I have also included data related to Internet Explorer 6 and Office 2003 (the installed browser and office suite on the Windows CMPC). In all cases, testing was completed with the Classmates on battery power, since they won't be tied to a power strip most of the time. A few final notes: with the exception of the stress test, all applications were launched and run alone; they were shut down before the next test was completed. For the stress test, all applications listed in the table (except Internet Explorer) were launched in succession, without waiting for the next to load; although not utterly scientific, it certainly provides a feeling for how well the OS compensates for the minimal RAM and Celeron processor. All applications were launched for the stress test while the 118MB file was copying from an HP USB Digital Drive (reading a 2GB SD card) to the desktop; the recorded time is for all applications to launch successfully (the file copy was not necessarily complete). Additionally, the Metasys unit was running OpenOffice 2.0, an iteration of OO.org not known for its optimization in terms of speed. All times were rounded up to the nearest second. Here are the results: