Disk Cache and file fragmentation
Posted by: JohnHell
Date: August 03, 2010 04:30AM

Hey, this is not spam, but let me make an introduction before we discuss it ;-)

The other day someone showed me the new features of Diskeeper. Because it had a new one very interesting: IntelliWrite.

Basically (my guess), what it does is to prevent fragmentation by monitoring I/Os to disk and avoid to write to disk until the file is completed and therefore ready to write. As any decent OS and file-systems manage files in the way that if the space is previously allocated, no fragmentation required, the fragmentation disappears.

This is useful for those applications (easy 90%) that don't or can't allocate file sizes because they grow without any predictable pattern.

Ok, here ended the introduction to basics on file-sytems ;) (NTFS and FAT32 as we use Windows).

What I told above was known by me by years but I didn't consider the new deals nowadays.

The browsers are using the disk continuously to write files to cache. But they are doing it wrong!

Ok, ok, is not wrong, but, why don't they allocate space for the downloaded file?

In the HTTP conversation, usually, not ever, is told by the server the file size, so the browser could easily allocate space and then write the data to disk avoiding fragmentation and gaining in overall performance, except in the cases the server doesn't tell the file size, but in total it would be still good.

Questions

I just searched a little and I found that, maybe, Firefox 4 (I don't know which GRE version) would have some similar features I'm talking about, but, till then...:

Do you know if there is some config in GRE/K-meleon to do this? (I think it hasn't, of course, but better ask),

or third party utility emulating Intelliwrite by Diskeeper as it can only be installed from XP SP3? (what left almost all of us aside)

Could be K-meleon programmed apart with this feature included?


I know that a way, now we have more and more RAM memory, could be to use 0 bytes of disk cache and use only memory cache but for me, is not a solution and is not the best, even if RAM I/Os is faster than disk I/Os; and also I want to keep cache files, as, for example, flash flv caching ;)



Edited 1 time(s). Last edit at 08/03/2010 04:32AM by JohnHell.

Re: Disk Cache and file fragmentation
Posted by: JohnHell
Date: July 16, 2012 09:35PM

This is a follow-up to this topic because I continue digging.

Well, tired of all this fragmentation that one of the causes is related to the disk cache wipe by mozilla products when they crash, I found that there is a reported bug on this topic, from... 2001!!!:

https://bugzilla.mozilla.org/show_bug.cgi?id=105843

As looks no one at mozilla would do something about disk cache behavior and, therefore, fragmentation. I started to "flirt" with RAM Disk, even I hate it.

Well, not exactly with RAM Disks but with a system like Google Earth cache. This is: a way to have solid cache file that helps to avoid fragmentation.

And beyond this the only thing I found was something that works like it and this is the freeware RAM Disk utility imdisk. It creates virtual disks, ram disk actually, that dump all its data to a file after unmount them (if you choose to do it).

I'm just starting to use it but, beyond I could imagine, looks that it hasn't a huge impact on RAM memory consumption, helps to avoid continuous I/O (input/output) to hard disk and is quite fast (it's a RAM Disk, after all).

Anyway, it helps quite a lot to file fragmentation (no fragmentation at all) and is what I was looking for.

Maybe in a few days I change and I drop using it, but here I left this message for your knowledge ;)

K-Meleon forum is powered by Phorum.