Coda File System

Re: some questions

From: Shafeeq Sinnamohideen <shafeeq_at_cs.cmu.edu>
Date: Wed, 8 Mar 2000 13:40:44 -0500 (EST)
On Tue, 7 Mar 2000, August [iso-8859-1] Hörandl wrote:

> 
> Hi,
> I am a teacher in a technical high school in vienna, austria
> 
> We are considering coda and linux. We have got about 1500 pupils
> and 200 teachers working on about 170 workstations

You should be aware that Coda is at best beta quality software. While it
works reasonably well in a well connected network, weak connectivity and
recovery after a network partition keep turning up new and intersting
bugs. While we try to fix bug reported as quickly as we can, it may
take some time as they may be difficult to reproduce or hard to find. 

For comparison, our installation has about 20 users and 40 workstations or
laptops. In any case, you should try Coda on a handful of machines to get
to understand it well before deciding. 

> I would like to use coda for the home dirs and most of the
> programs (/usr and /opt); the server to be is a piii 450 with
> 256mb ram and fast disks (~ 30 GB)
> 
> - rvm: this should be 4% of total data area - but the biggest
>   choice offered is 315M which is good for 8GB Data 
>   any ideas for 30G ?
>   or should i get a farm of "little" servers ?

A server process is limited to 2GB of RVM because that is the largest
continuous block of VM available on 32 bit machines. It does require some
amount of custom configuration to accomplish this; Phil Nelson is working
on making large servers easier to configure. 

A farm of small servers would probably be better, as it would allow you
to do replication and balance loads better and perform better as well.

> - if i create the volumes on one server first, is there
>   a possibility to replicate them as soon i get more servers ?

You can expand a Volume Storage Group  at a later date, but that affects
all volumes stored on that VSG and clients won't notice the change until
they restart, which could lead to the new server missing updates until all
clients have restarted. Resolution should automatically fix the missed
updates unless the old server also goes down. 

> - the 1700 users are organised in 5 big sections (different
>   types of education, teachers and school administration) 
>   the pupils are grouped into classes with about 25 pupils each
> 
>   any ideas how to organise the volumes ? i have all the user data
>   in a database and want to create all the useraccounts and volumes
>   in a batch job - so complicated directory structures should be 
>   no problem
> 

A volume can be mounted anywhere in a directory heirarchy, and can have up
to a 32 character name. Typically, the volume name is somewhat related to
its purpose and where it is located. Also typically each user and
project have their own volumes and additional volumes mounted under it if
they have a large amount of data or date with different requirements for
replication. While there is no limit on the size of a volume, keeping
volumes to around 50mb makes it easier to balance load, disk space, and
the amount of time a resolution takes after a server failure.  


I hope this helps, 
Shafeeq
Received on 2000-03-08 13:40:58