SCSI to IDE bridges...

[Chuck Guzis]

On 23 Sep 2008 at 23:50, der Mouse wrote:

> For example, if my "8G" microdrive were 8G of CF instead, and had (say)
> a write cycle count of 250K writes, that would mean I shouldn't expect
> it to wear out until after I've written approximately 250K * 8G, which
> is roughly 2PB.  I don't know what microdrive throughput is like (never
> measured it on my z50), but at the approximately 7M/s I'm seeing right
> now to a "real" disk I'm doing the write phase of a write-read-compare
> test over, it takes over 9.7 years to transfer 2PB - and that means
> 7M/s sustained for the whole 9.7+ years.

My understanding is the cycle count relates to writing to a given 
block.  For use as a swap device or other heavy rewrite-in-place 
device, the 250K writes could be used up in a matter of weeks.  
Anecdotal evidence I've run into on the web seems to bear this out.  

I've toyed with the idea of setting up a Linux box with a decent-size 
(512M) hunk of RAM,turning off swap and using it as a network 
mailserver with CF instead of IDE disk to conserve power.  I'm not 
sure how long the CF would last.  Assuming, say, 1000 emails 
(including spam) a day worst-case, I might see one fail in less than 
a year, depending upon where the writes landed.  Leveling would help, 
of course, but do CF cards auto-level, or does that responsibility 
fall upon the system using them?

Does CF as working storage even approach the reliability of a floppy? 
I can recall running 8" floppy read-write tests for many days at a 
stretch.

Cheers,
Chuck