9fans archive / 1994 / 11 / 14 /    prev next

From: Alberto Nava beto@pla...
Subject: yet another benchmark
Date: Fri, 18 Nov 1994 14:10:41 -0500

In <94Nov17.000230est.12684@gal...>
	Scott Schwartz <schwartz@gal...> wrote:

> Larry McVoy posted a filesystem benchmark on comp.unix.bsd today,
> Message-ID: <3ae05t$afs@fid...>.  I'm not sure what it's
> intended to prove, but since I'm a sucker for benchmarks I ran it
> anyway. :-)
> 
> Here are the results and the code (tweaked to compile).  Maybe someone
> with more interesting hardware than an SS1 cpu server and SS2 file
> server can share their results.  
> 
> cpu% k.out
> 0k	500	36	63
> 1k	500	19	50
> 4k	500	15	56
> 10k	500	11	56
> cpu% ramfs
> cpu% cd /tmp
> cpu% $home/k.out
> 0k	500	42	100
> 1k	500	36	100
> 4k	500	28	125
> 10k	500	11	17
> 
> 

We don't have a great hardware, but we are my results.

#
# terminal: SPARC ELC 
# file server: Sparc ELC
# cpu server: Sparc ELC
#

term% k.out
0k	500	26	45
1k	500	13	38
4k	500	11	29
10k	500	8	26

cpu% k.out
0k	500	26	63
1k	500	13	38
4k	500	12	28
10k	500	10	28

#
# with ramfs
#

term% $home/src/ben/k.out
0k	500	56	125
1k	500	50	100
4k	500	42	125
10k	500	9	24	

cpu% $home/src/ben/k.out 
0k	500	63	167			  
1k	500	56	125
4k	500	42	167
10k	500	28	167

The result for our cpu server running ramfs are a little better. Probably
becuase it has more RAM.