Revisiting Linux Part 1: A Look at Ubuntu 8.04
by Ryan Smith on August 26, 2009 12:00 AM EST- Posted in
- Linux
File/Networking Performance
Finally we have file and networking performance. As Ubuntu uses a different file system (ext3 versus Windows’ NTFS) there’s the potential for some significant differences here.
Starting with file performance, we will be using a collection of roughly 1500 files, totaling 380MB.
Even after SP1, file performance has long been a thron in the side of Vista. Here it manages an embarrassing loss to Ubuntu, taking over 50% longer to make a copy of the same folder.
Looking at ZIP compression times, it’s an even larger gap. Vista needs 75% longer to compress the same folder. While compression can be CPU bound, looking at our data this specific test is largely I/O bound. We’ve already established at Windows’ built-in ZIP abilities are pretty bad, but we’ve never figured out why this is.
Decompression is even worse for Windows. It takes nearly 4 times as long to decompress the same archive. It’s not even a contest – Ubuntu wins, if only because it’s the only competent operating system out of the two.
Meanwhile in our network copy tests, we are copying that folder to a server running Windows Server 2003. This gives Windows an advantage since we’re using SMB, but since SMB is the predominant protocol for consumer file server gear, it’s a fair test of such use.
Here it’s nearly a dead heat. Both Ubuntu and Vista need just as long to copy our file collection to the server, meanwhile Windows needs a bit less time to copy that same folder off of the server.
Switching gears, using a 2.6GB ISO we see a clear performance difference. In both copying to and from the server, Ubuntu needs at least 50% longer. Since this test isn’t using a lot of CPU time, our best guess is that Windows is doing some buffering that Ubuntu doesn’t get to do. The transfer rates for Linux are below what the hard drives on either end can manage.
Ultimately for users with lots of local storage, Ubuntu appears to outshine Vista. But for users with lots of remote storage (e.g. a NAS), Vista outshines Ubuntu.
Finally we have the amount of time it takes to start up each operating system, another disk-bound test. Vista is not something I would consider particularly speedy, so I’m a bit surprised that Ubuntu did not manage to outperform it here. The 2 second difference is measurable, but small enough that it won’t make any real impact.
Facebook
Twitter
RSS
195 Comments
View All Comments
jasperjones - Wednesday, August 26, 2009 - link
I second most of Fox5's suggestion.1.) I've been completely ignorant of software development on Windows over the last few years. Comparison of MS Visual Studio vs Eclipse or vs Netbeans/Sun Studio? How fast are CLI C++ apps on Windows vs. Linux? Perhaps using both GNU and Intel C++ Compiler toolchains on Linux. And possibly MS Visual C++ and Intel Visual C++ on Windows.
Perhaps less esoteric, 2.) instead of benching SMB/CIFS on Windows vs Samba on *nix, bench something *nix native such as scp/sftp or nfs. Netperf.
3.) Number-crunching stuff. I guess this is sort of similar to running at least a few synthetic benches. LINPACK or some other test that uses BLAS or LAPACK, tests that use FFTW. Maybe even SPEC (I wouldn't expect any exciting results here, though, or are there?)
Eeqmcsq - Wednesday, August 26, 2009 - link
Are you looking for benchmarks in Windows vs Ubuntu with the same hardware? Or benchmarks in different CPUs/motherboards/etc with the same Ubuntu?Ryan Smith - Wednesday, August 26, 2009 - link
Cross-platform. There's no problem coming up with Linux-only benchmarks for hardware.Eeqmcsq - Wednesday, August 26, 2009 - link
I have a question about your benchmarks that involve files, such as copying and zipping. When you run your benchmarks, do you run them multiple times and then get an average? I ask that because I have learned that in Linux, files get cached into memory, so subsequent runs will appear faster. I suspect the same thing happens in Windows. Do you take that into account by clearing cached memory before each run?Ryan Smith - Wednesday, August 26, 2009 - link
We reboot between runs to avoid cache issues (and in the case of Windows, wait for it to finish filling the SuperFetch cache).fri2219 - Wednesday, August 26, 2009 - link
I heard Sony is coming out with this thing they call a Walkman.You should review that next!
StuckMojo - Wednesday, August 26, 2009 - link
ROFL!Fox5 - Wednesday, August 26, 2009 - link
The LTS is really for the same types of people that avoid grabbing the latest MS service pack. IE, anyone who's still running Windows XP SP2 with IE6. Do that comparison and see how they compare.Ubuntu is little more than a tight integration of many well-tested packages, there's no reason to go with ubuntu's LTS when everything else already goes through it's own extensive testing. Given how quickly open source software advances, I'd say the LTS is probably less stable than the most up to date versions, and certainly far behind on usability.
You want the equivalent of Ubuntu's LTS in Windows? It most closely matches the progression that the Windows server versions follow.
Ryan Smith - Wednesday, August 26, 2009 - link
To put things in perspective, 8.04 was released shortly after Vista SP1 and XP SP3 were. So Hardy vs. XP SP2 (a 4 year old SP) is a pretty poor comparison.You'll see an up to date comparison in part 2 when we look at 9.04.
awaken688 - Wednesday, August 26, 2009 - link
I'm glad you did this article. It really has been something I think about. I'm ready to read your Part II. As others have mentioned, I have a couple of other articles that would be great.1) The comparison of the various versions as mentioned. SuSe, Ubuntu 9.04, BSD, etc...
2) Someone mentioned VirtualBox. I'd love to hear more about this including a detailed setup for the normal user. I'd love to be able to surf while in Linux, but able to play games in Windows and keep them separate for added security.
Thanks for the article! Hope to see one or both of the ideas mentioned above covered.