Coda File System

Performance and production environment

From: Daniel Savard <dsavard_at_cids.ca>
Date: 03 Mar 2002 01:34:56 -0500
I was interested in Coda since a long time. However, I never found
enough time to experiment with it and implement it.

I am now ready to jump in the bandwagon. However, since I haven't keep
an eye on the project for much time. I would like to know about two
things:

1) In the CODA How-to, which is somewhat old, there is a uppercase
warning saying CODA is not ready for production environments. After
searching a reading a little bit from the mailing lists, it seems this
is not true after two extra years of development. If CODA is ready for
production, I think the How-to document should be changed or the warning
should be slightly modified in order to not discourage those who are
interested to implement it.

2) It seems, still from the mailing lists, CODA is very good at
read-only filesystems and not so good at read-write filesystems. Is
there anyone who has run some kind of benchmarks to give us a better
idea of what it means, not good at read-write filesystems? How it
compares to NFS. I suppose the RVM log is a bottleneck, so is there any
significant improvements with some specific disks, filesystems layouts?
Is there any dependancy between bad performance at read-write and the
number of replicas? In short, someone can shed some light on this
assumption and give some hints and tips at tuning or planning the Coda
implementation?

I think Coda is having all the features I am looking from a network
filesystem. But I would like to make sure it will meet the requirements
before investing significant time in its implementation in my
environment.

-- 

=======================================
Daniel Savard
Consultation Informatique Daniel Savard
551, boul. Sir-Wilfrid-Laurier
suite 105
Canada J3G 4J1

Internet: dsavard_at_cids.ca
=======================================
Received on 2002-03-03 01:35:07
Binary file ./codalist-2002/4106.html matches