From: agonta 
Subject: Re: Standby vs. Cluster
Date: 2000/07/31
Message-ID: #1/1
Content-Transfer-Encoding: 7bit
References: 
Content-Type: text/plain; charset=us-ascii
Organization: @Home Network
MIME-Version: 1.0
NNTP-Posting-Date: Sun, 30 Jul 2000 20:18:13 PDT
Newsgroups: fido.novell

Tawn,

I have worked with many, many products. I've worked with
Vinca Standby server, doing one to one and one to many
configuration. I have worked with SFT-III and I have
worked not with mirroring, but file replication, 
with a program called "double-take" from nsisoftware.com
I have likewise played with clustering in lab.

Here's my opinions:

a) SFT-III vs Vinca: Winner: Vinca.
Why? Because it mirrors only the disk, not the RAM,
and easier to work with, and also allows for
separate servers to exist, not simply I/O engines
that are part of the same "logical" server.

Thus, you can run separate Third Party NLM's,
such as ArcSErve, NAV, Inoculan, etc, without
the sensitivity of SFT-III. You also gain benefit
of "snap shot server", and a "utility server.  
Please refer to Vinca site for more info.

b) File replication via SFT-III vs Vinca:

For offsite disaster recovery, it's fine, such
as NSISOFTWARE.Com. However, it's not mirroring
but simply a "constantly running" incremental copy
job that only transfers file deltas, not full files.
However, it's slow compare to SFT-III, or Vinca,
and is a software solution, not hardware solution.
However, you'll need to create alias objects to
the server's actual volumes. Thus, people will login
into the alias, not the real server in the login
script, and you can point the alias to one of 2 or
3 servers. For example, suppose the server holding
date is FS-1/VOL1: and the replicate server FS-2/VOL1:
thus, you'll need to put different hardware online, but
people don't login in contextually to a server by name,
but to the alias and different physical hardware could
be the "live" server. Good solution for offiste disaster
recovery, but not hardware failure onsite.

c) Server clustering: 
Combination hardware/software solution. Biggest
weakness: single point of failure. Typical scenario
involves large disk subsystem hooked up via fiber switch
to one of several NEtware 5.1 boxes. Each boxes, if it fails,
passes the disk subsystem to another "live box". This
in turn means that data is always accessible, from at
least one box. Typical scenario involves 5 boxes.
Solution is poor because it prevents failure of a server,
but not failure of disk subsystem of fiber switch hooking
servers to this box. A better solution is the "one to many"
solution of Vinca. One server runs 4 nic cards that are dedicated
links to 3 other servers. It has enough disk space to equal
the partition of each of the 3 other servers. It has RAM
equal to largest of the boxes. This server can
"stand-in" for any of these 3, and is a complete failsafe,
hardware solution, involving complete redudancy. Solution
is tricky, and complex to manage. One-to-one via Vinca,
while always requiring duplication of hardware, is
expensive ,but cleaner.

               Hope this helped.

                             Alex

Tawn McDowell wrote:
> 
> We are looking to upgrade our SFT-III server to Netware 5.  Does anyone know
> if Standby server offers an automatic, simultaneous switch over in the event
> the primary server goes down or is there a delay for the clients?
> 
> Thank you,
> Tawn McDowell