Reducing the size of the ports tree (brainstorm v2)

Chris H bsd-lists at bsdforge.com
Wed Nov 5 04:24:56 UTC 2014


On Tue, 4 Nov 2014 20:29:44 -0700 (MST) Warren Block <wblock at wonkity.com> wrote

> On Tue, 4 Nov 2014, Chris H wrote:
> 
> > On Tue, 4 Nov 2014 16:16:09 -0700 (MST) Warren Block <wblock at wonkity.com>
> > wrote >
> >> On Tue, 4 Nov 2014, Chris H wrote:
> >>
> >>> gpart(8) -a gives you what you need. If it's truly as bad as all that,
> >>> mounting the ports tree on a 512k aligned slice will reduce the "slack"
> >>> you appear to be referring to. zfs(8) also has this ability.
> >>
> >> Not alignment, but filesystem block size.  But that can only be set for
> >> an entire filesystem, and it's a tradeoff.
> >
> > Quite true. Which was meant to be my point.
> > Meaning that the ports tree could then be mounted where ever was
> > deemed convenient, and wouldn't carry the "slack" it does on a
> > 4k boundary. Maybe even on a removable SSD?
> 
> I thought that block suballocation was a thing on most modern 
> filesystems.  There would still be an extra seek or several to locate 
> the small sub-blocks inside a full block, but it should make space usage 
> with small files more efficient.  But I don't know what either UFS or 
> ZFS does for that.
Difficult to tell for sure. I haven't examined the [UFS/ZFS] source
to know for sure. Be valuable info. :)
OTOH I only mentioned utilizing a smaller boundary, as I felt it
was a reasonable solution related to size issue mentioned.
I have just about enough spares laying about, to do some comparison/
benchmarking on UFS v ZFS v 4k v 512b. If I get a chance this week.
I'm going to give it a go, and see if I can extrapolate useful data.

--Chris




More information about the freebsd-ports mailing list