Wednesday, 09 August 2006

The beauty of having your hardware arrive in installments is you can do all sorts of fooling around with lots of storage that you wouldn't otherwise do because you'd be falling over yourself to get it all up and running yesterday.

So as I impatiently wait for my TV tuner card to arrive, I've been doing some "benchmarking" to see how various different configurations perform. One of the things I wanted to find out was how much overhead did LVM give you.

So I took one of the four disks in my non-redundant array of Bloody Huge Storage, and ran bonnie++ on it as a straight JFS volume on /dev/sda and as a logical volume in a volume group that used all of the logical extents provided by making /dev/sda into a physical volume.

Again, I ran bonnie++ as:

bonnie++ -d /mnt -s 2G -n 1024 -u nobody:nogroup

So here's the results of running it on a JFS filesystem straight on /dev/sda:

                    -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
minotaur         2G 13712  95 18561  21  9152  11 13874  94 20446  18 142.2   1
                    ------Sequential Create------ --------Random Create--------
                    -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
              files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
               1024  5999  37  7125  19  1604  12   616  12   168   0    61   1

And here's the results of running it on a JFS filesystem on a logical volume that used all of the free logical extents provided by a volume group consisting of a physical volume made from /dev/sda

Version  1.03       ------Sequential Output------ --Sequential Input- --Random-
                    -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
minotaur         2G 13790  96 18566  21  9132  11 14202  96 20628  18 146.3   1
                    ------Sequential Create------ --------Random Create--------
                    -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
              files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
               1024  5989  37  7056  20  1605  12   611  12   167   0    61   1

Yup, that's right. It was actually slightly faster in a few cases. I can't explain that one. But it certainly goes to show that there is no noticeable performance impact of using LVM. Nice.

Film at 11

My friend Andrew brought this article in the Sydney Morning Herald to my attention today.

I was initially mildly incensed by it, then someone at work pointed out the author's Wikipedia entry, which led onto this Bulletin article on her. So it seems she's just a professional right-wing shit-stirrer. This other article she wrote managed to stir up some ire amongst the cycling community.

Still, I am glad to see that An Inconvenient Truth is coming to Australia. Well worth the watch. Now I need to go find where Who Killed the Electric Car is screening...