SSD vs. HD failure
Posted Jun 19, 2009 4:43 UTC (Fri) by djao
Parent article: What ever happened to chunkfs?
There's a good bit of FUD in the middle of the article when it talks about SSDs. Of course, it is certainly true that overused individual disk sectors in SSDs tend to fail. However, a fairly large percentage of these failures occur on write, meaning that you know something went wrong, but your old data is still there. In such situations, the corruption resilience properties of chunkfs are not really that useful.
If we now turn our attention to regular hard drives with mechanical platters, one of the common ways that such drives can fail involves the entire drive dying at once. (This can happen to SSDs too, but much more rarely.) Chunkfs won't really help very much in this case either. I should also add that mechanical failure of mechanical drives is more common than SSD failure. Moreover, when a mechanical drive fails for whatever reason, it almost always fails on read instead of (or in addition to) on write, meaning that you've lost old data. So, even though SSDs aren't perfect, they're still better than the alternatives if your goal is reliability. And even though the chunkfs concept is handy for a large class of disk failures, you're still better off with ext3 on an SSD than chunkfs (or any FS really) on a mechanical drive.
to post comments)