Re: [Gnu-arch-users] Some issues

[James Blackwell]

I don't know who:
>> But it's so much cheaper to read the data from a single, larger file,
>> than from lots of different ones.
>

Andrew Suffield:
> This is essentially wrong on all the interesting filesystems. It's
> equivalent on most of them, and a little slower to use one large file
> on NTFS and HFS+.

Nonsense. You should know better than this.

As far as I'm aware, just about every hard drive out there has 512 byte
sectors. Arch ids are somewhere around 70 bytes each. That means that
one can fit 7 arch-ids in one sector. On the first sector you load,
you're looking at 8 ms.

What's more, all of the interesting filesystems are pretty good at laying
files out in a nice order, which keeps access time low.

But let's forget that, and go back to the 7 arch-ids in one sector
analogy. To stat those 7 archid files, you're looking at an average
cumulative latency of 56 ms, for a net loss of 49 ms.. 

If you then extend that to 200 files, you're looking at 1.633 seconds
(200 files / 6 extra files statted for inodes * 56 ms). If you're
looking at a tree with 10K arch-ids, you're now up to 81 seconds. 

What's more, but you loose even more than that (relatively speaking),
thanks to the wonders of things such as DMA.


-- 
James Blackwell          Try something fun: For the next 24 hours, give
Smile more!              each person you meet a compliment!

GnuPG (ID 06357400) AAE4 8C76 58DA 5902 761D  247A 8A55 DA73 0635 7400