On Thu, Jan 22, 2004 at 10:51:00AM -0600, Richard Smith wrote: > Our most common > scenario for failure is to lose disk block #0, which, of course, is the > boot block for our OS.
Well that would be a drag, that block #0 is nice to have around. At least you don't have head crashes, my hope is that CF is at least as reliable as HD given limited controlled writes. For my application I'm looking at logs that grow about 4MB/month(log every 5 minutes about 350 bytes). They are erased every 3 months. Then I want to save state every 5 minutes, which is the bigger problem area. Right now I save 30KB every 5 minutes. I haven't optimized it yet for size, but I should be able to reduce the size of this by at least 10 fold. Which will help. But I'm also concerned about proper wear leveling given this constant re-writing of the same data file. Given what SanDisk says about wear leveling being localized to smaller blocks, I want to ensure that the file is being moved around enough to guarentee good wear leveling. I may be able to just dd the raw partition over and look at it to get a feel if the status file is moving around or staying at one spot. If it's staying in one spot, I could convert it to writing a large set or appended/log to move it around some. I got a few IDE/Flash card things from www.linnix.com. They have what I'll call a home-made socket(wires straight to CF with epoxy formed around it). Definitely not military spec, but good for those inexpensive toy projects.
