Note to self: Try out the data=journal option for ext3 and see if it has any effect on performance in desktop systems. This link, while old, suggests heavy read-write loads might see an improvement for read access times.
I have my doubts, personally. It sounds like that was a possibility seven years ago under the 2.4 kernel, and probably with server loads on server machines. I just don’t expect a 550Mhz Celeron running a lightweight desktop to see an improvement in start and performance times by adding to the filesystem load.
Thanks to issueperson just the same, for pointing it out here.
It does in fact improve interactive performance, I’ve tried it myself on both arch and ubuntu with a recent computer (core 2, 2GB). However, as you expect raw I/O performance is NOT increased, instead it is drastically decreased, to around 1/2 of its normal value. However, for most tasks this difference is overshadowed by the interactive benefit obtained. I didn’t notice any significant change in boot times, maybe 1-2 seconds, but file copying was much slower. On the other hand, no matter how much file activity was going on applications remained completely responsive, unlike with normal ext3. Its worth a try if you don’t frequently need to work with large files.
That’s good to know. I’ll give it a try this weekend and see what happens. It might be oversimplifying things, but personally I’m only interested in desktop performance, for simple things like starting up and opening a browser. If the benefits appear only at particular times or during particular workloads, I’m not as interested.
Pingback: ext3 with data=journal results « Motho ke motho ka botho