🚀 go-pugleaf

In <1993Jan8.081350.3453@ica.philips.nl> adrie@ica.philips.nl (Adrie Koolen) writes:

| In article <1943@niktow.canisius.edu> pavlov@niktow.canisius.edu (Greg Pavlov) writes:
| >                      SGI Indigo R4000     SUN 4/75 SPARC 2
| >
| > *Sequential disk test       1707KB/sec            602KB/sec
| > *"Random" disk test         1219KB/sec            328KB/sec
| >   (*8192-byte blocks)
| 
| I'll examine the result of the random disk test on the Indigo.
| Using 8KB blocks, transfering one block at 4MB/s takes some 2ms.
| The Indigo reads 1219/8 = 152 blocks per second, i.e. 6.6ms per
| block. This means that the mean access time of the disk used by
| the Indigo is some 4.6ms, where I assume zero software overhead
| of the operating system.

If the file is small enough that the whole thing fits in memory,
we may just do a better job of caching than the Sun system.
Sun improved their performance for some of this stuff in 4.1.X.

We may also be doing a better job of keeping the file contiguous
on disks (or the filesystem may have been less fragmented in the
first place) leading to shorter seeks.

That is, I really don't think that we are hitting the disk for
every request.
--
Let no one tell me that silence gives consent,  |   Dave Olson
because whoever is silent dissents.             |   Silicon Graphics, Inc.
    Maria Isabel Barreno                        |   olson@sgi.com