| 1 |
Michał Górny posted on Fri, 17 Jan 2014 17:27:30 +0100 as excerpted: |
| 2 |
|
| 3 |
> Now some numbers. I did some tests 'converting' late gx86 daily tarballs |
| 4 |
> to squashfs. I've used squashfs 4.2 with LZO compression since it's |
| 5 |
> quite good and very fast. |
| 6 |
> |
| 7 |
> 96M portage-20140108.sqfs |
| 8 |
[...] |
| 9 |
> 97M portage-20140114.sqfs |
| 10 |
> 97M portage-20140115.sqfs |
| 11 |
> |
| 12 |
> For deltas [...] |
| 13 |
> |
| 14 |
> 4,9M portage-20140108.sqfs-portage-20140109.sqfs.vcdiff.djw |
| 15 |
> 6,3M portage-20140109.sqfs-portage-20140110.sqfs.vcdiff.djw |
| 16 |
[...] |
| 17 |
> 8,5M portage-20140114.sqfs-portage-20140115.sqfs.vcdiff.djw |
| 18 |
> |
| 19 |
> As you can see, the deltas are quite large compared to the actual |
| 20 |
> changes. However, we could have expected that since we're diffing a |
| 21 |
> compressed filesystem. What's important, however, is that applying it |
| 22 |
> takes ~2.5 second on my 2 GHz Athlon64. |
| 23 |
|
| 24 |
And... eyeballing a 6 MiB average, diffs are ~1/16 the full squashfs |
| 25 |
size, perhaps a bit larger. So people updating once a week or even about |
| 26 |
every 10 days would see a bandwidth savings, provided the sync script was |
| 27 |
intelligent enough to apply updates serially. |
| 28 |
|
| 29 |
The breakover point would be roughly an update every two weeks, or twice |
| 30 |
a month, at which point just downloading a new full squashfs would be |
| 31 |
easier, at about the same bandwidth. |
| 32 |
|
| 33 |
> What do you think? |
| 34 |
|
| 35 |
How does this, particularly the metadata cache, interact with overlays? |
| 36 |
That's /my/ big question. |
| 37 |
|
| 38 |
-- |
| 39 |
Duncan - List replies preferred. No HTML msgs. |
| 40 |
"Every nonfree program has a lord, a master -- |
| 41 |
and if you use the program, he is your master." Richard Stallman |
Archives