Cleaning out some old research papers I came again across the proposal
for "space maps" in ext4 [1]. What happened to that? Was it rejected
due to too much risk or other technical reasons? The benefits sound
appealing, so I was wondering if there's something fundamentally wrong
with the suggested approach.
Holger
[1] https://www.kernel.org/doc/ols/2010/ols2010-pages-121-132.pdf
On Mon, Nov 24, 2014 at 12:43:51PM +0000, Holger Hoffst?tte wrote:
>
> Cleaning out some old research papers I came again across the proposal
> for "space maps" in ext4 [1]. What happened to that? Was it rejected
> due to too much risk or other technical reasons? The benefits sound
> appealing, so I was wondering if there's something fundamentally wrong
> with the suggested approach.
No one ever submitted patches. The other issue would be the 71 patents
which Jeff Bonwick claimed in his blog that Sun filed on technologies
relating ZFS. I wouldn't know whether there are any patents that read
on space maps, because like many companies (due to the United States'
dysfunctional patent system), my employer doesn't allow me to look at
patents --- and so I don't want to know (see previous comments about
the US's dysfunctional patent system). However, if anyone knows of
space maps being used in some file system before ZFS, that would
certainly be interesting and something that I _would_ want to know.
There is some thinking about using an in-memory AVL tree to track free
space, and demand-loading it from the bitmaps, but no one has had the
time to implement it to date.
Cheers,
- Ted
On 11/24/14 9:33 AM, Theodore Ts'o wrote:
> On Mon, Nov 24, 2014 at 12:43:51PM +0000, Holger Hoffst?tte wrote:
>>
>> Cleaning out some old research papers I came again across the proposal
>> for "space maps" in ext4 [1]. What happened to that? Was it rejected
>> due to too much risk or other technical reasons? The benefits sound
>> appealing, so I was wondering if there's something fundamentally wrong
>> with the suggested approach.
>
> No one ever submitted patches.
Was just checking that too. I'm always mystified by the folks who take the
time to write & submit & present papers, but don't submit patches.
-Eric
On Mon, 24 Nov 2014 09:38:27 -0600, Eric Sandeen wrote:
> On 11/24/14 9:33 AM, Theodore Ts'o wrote:
>> On Mon, Nov 24, 2014 at 12:43:51PM +0000, Holger Hoffstätte wrote:
>>>
>>> Cleaning out some old research papers I came again across the proposal
>>> for "space maps" in ext4 [1]. What happened to that? Was it rejected
>>> due to too much risk or other technical reasons? The benefits sound
>>> appealing, so I was wondering if there's something fundamentally wrong
>>> with the suggested approach.
>>
>> No one ever submitted patches.
>
> Was just checking that too. I'm always mystified by the folks who take the
> time to write & submit & present papers, but don't submit patches.
Thanks - that's all very sad. The author has a homepage at
http://saurabh.io/ so maybe someone with a sufficiently large army of
lawyers can contact him? I couldn't find any code right now either.
As for the patents - I'm outside the US and the paper explicitly mentions
using Linux' built-in RBTrees instead of AVL and a few othe minor
differences to ZFS, mostly relating to persisting the maps.
Holger