Well, I got venus outfitted with private mapping stuff ... and here are the
results of venus running with a 1Mb cache: (for startup times...)

Machine: (Coda machine) pentium (speed?), NetBSD-1.4.1, 64 meg memory, IDE disk.

Anonymous mapping:
Coda Venus, version 5.3.3

Date: Fri 12/03/99

11:38:52 /usr/coda/LOG validated at size 0x18ee200
11:38:52 /usr/coda/DATA validated at size 0x63b7080
11:38:52 loading recoverable store
11:39:45 Last init was Fri Dec  3 11:18:36 1999
11:39:45 Last shutdown was clean
11:39:47 starting VSGDB scan
11:39:47 	1 vsg entries in table
11:39:47 	0 vsg entries on free-list
11:39:48 starting VDB scan
11:39:48 	2 vol entries in table (0 MLEs)
11:39:48 	0 vol entries on free-list (0 MLEs)
11:39:49 starting FSDB scan (41666, 1000000) (25, 75, 4)
11:39:49 	0 cache files in table (0 blocks)
11:39:54 	41666 cache files on free-list
11:39:55 starting HDB scan
11:39:55 	0 hdb entries in table
11:39:55 	0 hdb entries on free-list
11:39:55 Getting Root Volume information...
11:39:56 Venus starting...

=> 1:04 start up time.

Private mapping.

Coda Venus, version 5.3.3

Date: Fri 12/03/99

11:40:34 /usr/coda/LOG validated at size 0x18ee200
11:40:34 /usr/coda/DATA validated at size 0x63b7080
11:40:35 loading recoverable store
11:40:35 Last init was Fri Dec  3 11:18:36 1999
11:40:35 Last shutdown was clean
11:40:35 starting VSGDB scan
11:40:35 	1 vsg entries in table
11:40:35 	0 vsg entries on free-list
11:40:35 starting VDB scan
11:40:35 	2 vol entries in table (0 MLEs)
11:40:35 	0 vol entries on free-list (0 MLEs)
11:40:36 starting FSDB scan (41666, 1000000) (25, 75, 4)
11:40:36 	0 cache files in table (0 blocks)
11:41:10 	41666 cache files on free-list
11:41:11 starting HDB scan
11:41:11 	0 hdb entries in table
11:41:11 	0 hdb entries on free-list
11:41:11 Getting Root Volume information...
11:41:11 Venus starting...

=> :37 startup time

In looking at them:  (times for each step)

            Anon       Priv
RVM         :53        :00
VSGDB       :01        :00
VDB         :01        :01
FSDB        :06        :35
HDB         :00        :00

RVM size is about 100 Mb

The -init stage took a very long time in the FSDB scan.  This looks like an
area where some effort could be put to reduce the startup of the FSDB, like
keeping a flag with the fso structure that says the file is created or not.
A lookup in the initial -init should be much faster than a create.

-- 
Phil Nelson                    NetBSD: http://www.netbsd.org
e-mail: [email protected]        Coda: http://www.coda.cs.cmu.edu
http://www.cs.wwu.edu/~phil