9fans archive / 1997 / 12 / 23 /    prev next

From: Kenji Arisawa kenji@nx
Subject: [9fans] Kfs file server
Date: Thu, 11 Dec 97 11:10:21 +0900

Hello!

Kfs file server of plan9 looks too slow.
The followings are the times I observed coping files of same  
size(5.66MB).
Experiments are done on i486DX2/66MHz machines.

cpu% time cp x y  # copy on kfs (ids disk). Block size is set 4096.
0.01u 2.11s 63.88r       cp x y

cpu% time cp x y  # copy on 9fs
0.03u 4.51s 24.64r       cp x y

ar$ time cp x y   # copy on unix system (4.3BSD) ide disk
        5.3 real         0.0 user         1.4 sys

C> copy x y       # copy on DOS ide disk
	10 sec (observed using my watch)

Coping on 9fs involves bi-directional transmission
through 10 base ethernet using 9P protocol.
The value is reasonable and may be efficient.
If we apply 100 base ethernet, this time will largely improved.

However why the result of kfs is so bad?
Caching problem?
Algorithm?

Can I improve this value?

I did this experiment in the hope that I can improve file access  
time
on plan9 using only kfs adding some security mechanism.

Kenji Arisawa
E-mail: arisawa@aic...