Subject: Re: more on mysql benchmark
To: None <tech-kern@NetBSD.org>
From: SODA Noriyuki <soda@sra.co.jp>
List: tech-kern
Date: 03/10/2005 19:27:37
>>>>> On Thu, 10 Mar 2005 10:24:22 +1100,
	Simon Burge <simonb@wasabisystems.com> said:

> Compare those times with a trimmed down kernel against a GENERIC kernel
> (6284 KB free at boot) for the vm tunables with the fastest time above.
> 
> time (real)                  vm.bufmem_hiwater vm.{anon,exec,file}{min,max}
> 8:55.57 44.2%  0+0k 3+24io 52780pf+1w  131072  {{10,80},{ 0,10},{ 0, 1}}

Simon looked at this further to see why 2.99.16 GENERIC is more than
4 times slower than 1.5X GENERIC, although both kernels provide nearly
same amount of free memory:
	   1.5X GENERIC: 6220KB
	2.99.16 GENERIC: 6284KB - 90KB (*1) = 6194KB

This was because file pages occupied 7% ~ 26% (296KB ~ 1076KB) of RAM
with the VM tuning parameters above:
			(pages)				     (%)
-----------------------------------------------------   --------------
 free	active	inactive	 anon	 exec	 file	anon exec file
  114	  600	  312		 1036	    2	  133    101   0   13
  198	  508	  321		 1011	    2	   74	  98   0    7
  201	  506	  320		  964	    2	  118	  94   0   11
  142	  570	  314		  870	    4	  269	  85   0   26
  134	  576	  315		  929	    8	  213	  91   1   21

BTW, it took 9:52.48 this time:
63.148u 182.001s 9:52.48 41.3%  0+0k 6+25io 56646pf+1w

(*1) dynamically allocated buffer cache: vm.bufmem=92160
     This value was measured by sysctl(8) at the time when the
     benchmark finished.

> And then compare that with a 1.5X GENERIC kernel (6220KB free):
> 
> 55.189u  20.468s  2:06.78 59.6% 0+0k 71+63io  5364pf+0w
--
soda