DragonFly BSD
DragonFly kernel List (threaded) for 2003-09
[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]

Ridiculous idea: Cache as ramdisk?

From: J. Kanowitz <jkanowitz@xxxxxxxx>
Date: 23 Sep 2003 12:59:16 GMT

I'm embarrassed to be posting this, since it's definitely a half-baked
sort of thing, but if ever there was a group of people who could take
it halfway seriously...

Basically, someone brought up the Amiga's recoverable RAM disk on
Slashdot, and the following thought ensued:
I don't want to burst your bubble... (Score:0)
by Anonymous Coward on Tuesday September 23, @08:25AM (#7032748)

. ..but this is sort of what modern OS's fat disk buffers are for. 
VFS caches make the entire filesystem a 'recoverable RAM drive' 
of sorts.

Only problem is, they assume all data is equally precious, and 
commit rather rapidly. What would be interesting would be something 
like a "blaze" command, which would twitch some kernel hook to 
ignore disk-commit timing for writes from a process.

So something like "blaze unzip bunchofjunk.zip" would proceed with 
no write delays, and the files would hang around in the cache for 
later reads, but I'd have no guarantee the junk files would ever 
commit to the FS unless I 'touch''d them afterwards or something.

Maybe "livedangerously" would be a better name.

I hope I haven't abused the terminology or concepts too badly there;
I'm thinking of wherever writes get buffered before a softupdates 
commit.  Obviously it'd suck to run completely OOM for careless use 
of the 'RAM disk,' MFS does a good job anyway, and I barely know what
I'm talking about...  But maybe the perspective can be good inspiration
for someone even if the idea itself might belong in the dustbin.

Newb-mindbending discussion is always appreciated. ;)

-Joe "Floid" Kanowitz 

[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]