Re: Proxy of large files

On Wed 22 Jan, 1997, David Luyer <luyer@ucs.uwa.edu.au> wrote:
>Under modern OS, Linux especially, I would hope for a performance
>improvement with the NOVM version. The buffering/caching in the OS is
>probably better and more efficient than keeping two or more sets of
>indexes for copies the same data, and possibly the same data in the
>cache/buffer and squid memory cache. It is for a similar reason that many

they're all good points, but each disk access still requires blocking
I/O. The original harvest cached ethos was, therefore, to read
it all from disk in as large chunks as possible and then send it out
from memory.

I've not yet studied what changes Duane has made for the NOVM version,
though a cursory glance doesn't show any mmap() usage and it looks like
it's "simple" reads. There's been quite a lot of discussion here about
async I/O and other methods. I feel that the NOVM code would work well
with that, or with the techniques that I believe Cached 2.0 uses.

-- jrg.
Received on Wed Jan 22 1997 - 03:22:04 MST