dtrace discuss - Nov 2006 - timestamp vs vtimestamp

Hello all,
 I''m looking for answers about a question that i did post on ZFS/UFS
discuss... without success, now i''m here to see your points. Any tips?
As you know, i''m making some "screencasts" about a few
solaris features. That screencasts is one point of many tests that i''m
making with solaris 10. Now, with some tests with dtrace, i have saw a
interesting point: the creat64 and unlink system calls are uniforms (times) on
ZFS, but have some strange times on UFS (standard deviation).
I have two simple questions:
a) The filesystem utilization can be the problem? I mean, the UFS filesystem is
86% in use, and the ZFS is almost empty.
b) Can that difference be structural? I mean, have a explanation on the design
of the two filesystems? If yes, can you quickly explain that to me (links to
opensolaris code)? So i can put it in the blog (screencasts).
Thanks very much for your time.

The url: http://www.posix.brte.com.br/blog/?page_id=30

Leal.
 
 
This message posted from opensolaris.org

Sorry,
 I did use that program to gather the times (timestamp):

#!/usr/sbin/dtrace -s
#pragma D option quiet
int x;

syscall::creat64:entry
/execname == "touch"/
{
 self->ENTRADA = timestamp;
 x = x + 1;
}

syscall::creat64:return
/self->ENTRADA/
{
 this->SAIDA = timestamp - self->ENTRADA;
 this->TIME = ( this->SAIDA / 1000 );
 printf("%10d : %10d milliseconds\n", x, this->TIME);
 self->ENTRADA = 0;
}
--------------------------------------------------------------------------------------------------
 
 
This message posted from opensolaris.org