A small doubt in coda replication process and file access

[dhiraj_c]

Hello,

I have a small doubt regarding the replication process in coda file system.

Suppose a volume is replicated between 2 or 3 servers and is mounted on
some X number of clients. Now if these clients are accessing different
files (mainly reading) from this volume simultaneously, does only a
particular server (scm) serves these files or any of the server having the
replicated copy serves the files?

The reason I am asking this question is that I want to know that in Coda
FS is there a possibility of any server serving a replicated volume
becoming a "hotspot"? Because if the value of X is 1000 or more than the
load in a particular server will be more and it may crash due to heavy
load.

Also does Coda FS follow some kind of load-balancing policy among
replicated servers?

regards,
Dhiraj Chawla


--
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.

[Jan Harkes]

On Thu, Oct 02, 2008 at 01:56:41AM +0530, dhiraj_c@... wrote:
> I have a small doubt regarding the replication process in coda file system.
>
> Suppose a volume is replicated between 2 or 3 servers and is mounted on
> some X number of clients. Now if these clients are accessing different
> files (mainly reading) from this volume simultaneously, does only a
> particular server (scm) serves these files or any of the server having the
> replicated copy serves the files?

Clients fetch the metadata/attribute information from all available
servers, this is how differences between the replicas are discovered.
As we fetch these attributes we track response times and use this to
adjust estimated network bandwidth and latency values.

The file contents is then fetched from the server that has the best
estimated bandwidth and lowest response time. If the estimates are
correct this tends to be the nearer, faster, or less loaded server.

Jan