Well, now that I seem to have Coda-4.6.1 working on PPC, I was considering
seting up my two machines as servers and starting to place some data on
them. (I want to have two machines with replicated volumes for data
redundancy, plus having a coherent filesystem which doesn't depend on
either machine being up all the time, as NFS would)

I re-initialized RVM with a 30 MB log and a 90 MB data partition, and a ~
1.8 GB /vicepa, and created a coda:root partition. I then ran bonnie, and
got the following results:

       -------Sequential Output-------- ---Sequential Input--  --Random--
       -Per Char- --Block--- -Rewrite-- -Per Char- --Block---  --Seeks---
    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
    10    77  1.8    90  0.9    85  1.6    95  1.5    96  0.4  37.0 1.7

What's going on here? Is this because I'm running the client on the same
machine as the server? Is there some strange deadlock going on? I'm using
a 2.1.115 vger cvs kernel (vger is a cvs archive with the most recent PPC
changes) and the coda-linux-2.6.2 venus client, and my patched coda-4.6.1
 server.

What kind of performance do you get with the in-house Coda servers at CMU?
(and what kind of machines are they?)

I also ran this from an old slow 486/33 client running 2.0.35, and to my
surprise, the results were better:

       -------Sequential Output-------- ---Sequential Input--  --Random--
       -Per Char- --Block--- -Rewrite-- -Per Char- --Block---  --Seeks---
    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
     2   208 29.1   246 12.9   139 13.7   203 36.1   279 18.1  61.5 18.6

--------------------------------------------------------------------------
| Troy Benjegerdes    |    troybenj_at_iastate.edu    |    hozer_at_drgw.net   |
|    Unix is user friendly... You just have to be friendly to it first.  |
| This message composed with 100% free software.    http://www.linux.org |
--------------------------------------------------------------------------