On Wed, Apr 27, 2005 at 08:26:15AM -0600, Patrick Walsh wrote:
> One thing that I'd also like to see -- maybe its place is on the wiki
> -- is a list of Coda's constraints. How many files in a directory can
> coda handle, what are the limitations when working with big files, etc.
Some of these I know, but there are probably many I don't even know.
volume names: 32 characters
file names: 256 characters
path names: 1024 characters
ACLs per directory: 20
- Don't know why I always thought it was 6
directory size: 256KB
- this doesn't easily translate to # of files, because it
depends on the average filename size, padding, space for the
file identifiers, etc.
a ballpark figure would be between 2048 and 4096 files.
volumes/server: 1024
replicas/volume: 8
users and groups: unknown, I tried several thousand
users/group: unknown, at least several thousand as well
- Here is a CVS commit message related to that,
2003/06/17 20:39:23 <[EMAIL PROTECTED]>
Caching the rwcdb handle across accesses. Time to import 5000 users
and a group containing all users went (on my laptop) from 21+ minutes
to 11.621 seconds. The 100,000 user test is taking a lot longer to
import but checking group membership as the import is still going
directories/volume: Depends on the size of the volume resolution log
- We use one 'ResolveNull' entry for each directory.
resolution log: Administrative limit
- By default a volume has 4096 entries, but this can be increased
with 'volutil setlogparms <volid> reson 4 logsize <n>'.
files/volume: Unknown
- not as simple as (files/directory * directories/volume) because
at some point we have trouble finding enough unfragmented RVM
memory to allocate the flat lookup table. I think RVM allocation
issues hit around 200,000 files/volume, but I could be wrong.
servers/realm: 253
- We need unique server-id's in the range [0:255]. But 0, 127, and
255 are 'reserved' to indicate 'no volume, replicated volume,
and error'.
rvm data/server: anywhere between 1GB and 3GB
- depends on where the OS loads it's shared libraries and such.
... probably many more
Jan
