Round-Robin with File Servers

[Guest]

Archived from groups: microsoft.public.win2000.dns (More info?)

I may be way off on this, but I thought I would ask anyway. I am trying to
design a network with 2 file servers for a client. What they want is to have
these 2 servers act as one. IE if one server goes down then the other kick in
and away they go. Each server will have the same files as they are going to
replicate on a schedule across gigabit cards on their own network.

My question is this: If I were to setup Round-Robin in the dns, will this be
a viable solution?

I know that RR balances the load between servers and amongst my research, it
works great for web and email servers. Would this be a solution for file
servers? I am thinking i have to lean towards clustering, no? If so, is there
a cheaper way to go then Clusters? What about NLB? Thanks all.

Kevin

 

[Guest]

Archived from groups: microsoft.public.win2000.dns (More info?)

"Kevin" <Kevin@discussions.microsoft.com> wrote in message
news:A8E586A4-7A2A-4D1A-84B8-601485B58C0D@microsoft.com...
> I may be way off on this, but I thought I would ask anyway. I am trying to
> design a network with 2 file servers for a client. What they want is to
have
> these 2 servers act as one. IE if one server goes down then the other kick
in
> and away they go. Each server will have the same files as they are going
to
> replicate on a schedule across gigabit cards on their own network.

> My question is this: If I were to setup Round-Robin in the dns, will this
be
> a viable solution?

Probably not -- there are some oddities about
file and print sharing using an alternative name
sometimes -- it might work somewhat but there
is a larger problem.

Round Robin schemes do NOT provide fault
tolerance.

If one server (of N servers) is down, 1/N clients
will fail to reach the resources.

You have two major choices that actually work.*

> I know that RR balances the load between servers and amongst my research,
it
> works great for web and email servers.

(Roughly) Balances Load -- but NOT fault tolerant.

> Would this be a solution for file
> servers? I am thinking i have to lean towards clustering, no? If so, is
there
> a cheaper way to go then Clusters? What about NLB? Thanks all.

*Workable solutions:

1) Server clusting (not NLB)
requires shared drive space (special hardware)

2) DFS
some delay for replicating files even if you use
'automatic replication'

DFS is much more in the spirit of round robin but
DOES provide fault tolerance and even automatic
replication.




--
Herb Martin


>
> Kevin