Coda File System

Re: file data to storage server mapping in coda

From: Jan Harkes <>
Date: Wed, 13 Oct 2010 00:57:41 -0400
On Sat, Oct 09, 2010 at 05:22:32PM +0530, srimugunthan dhandapani wrote:
> the algorithm used to map a file's data to the storage servers). I
> couldn't  obtain  the information  about coda's data distribution
> algorithm from the coda papers. It will be helpful if somebody can

It is discussed in several of the papers, but the most concise
description is probably in 'The Evolution of Coda', which for some
reason doesn't seem to be listed on the page of Coda papers.

> give some helpful pointers about which files in the source code that i
> have  to look into for understanding how data distribution is done in
> coda.

Pointing at the actual source files is actually somewhat hard because
it is a somewhat fundamental concept in the design of the client. A high
level overview is, files on the server are stored in/associated with
logical volumes and when these volumes are created the administrator
decides which servers will be used as replicas for that particular
volume. When new file data is written to the servers it will be sent to
the set of volume replicas that is reachable (the Available Volume
Storage Group, or AVSG). When a file is read the attributes and version
information is retrieved from all available volume replicas, if any
replica is out of data resolution is used to propagate the new data to
the stale replica. Once all replicas are considered to be in sync the
file contents is retrieved from any one of the replicas, typically the
one stored on a server that historically responded quickest to requests.

Received on 2010-10-13 00:58:05