Text 4114, 60 rader
Skriven 2005-05-15 01:06:52 av Ellen K. (1:379/45)
   Kommentar till text 4110 av Geo (1:379/45)
Ärende: Re: 1 4 monte
=====================
From: Ellen K. <72322.enno.esspeayem.1016@compuserve.com>

On Sat, 14 May 2005 02:05:58 -0400, "Geo" <georger@nls.net> wrote in message
<428594f6$1@w3.nls.net>:

>"Ellen K" <Ellen.K@harborwebs.com> wrote in message
>news:b08485.ad8113@harborwebs.com...
>
>> Books Online says to implement failover clustering Windows Clustering has
>to be
>> set up, that's why I asked about it.
>
>That's the part I was thinking of and as I recall it's fairly expensive.
>
>> He mentioned that in other companies he implemented replication and then
>IT
>> would switch the users to the standby box in case of failure.  Is this one
>of
>> the "better solutions" you mention?  And/or could you list the ones you
>like
>> the best?
>
>I've never done it but it's possible to share a disk setup between machines,
>that way the spare machine just watches for the primary to fail and when
>that happens it picks up. The shared disks are raid5, usually a SAN or
>something like that.

That seems pretty slick.
>
>There is also a fallover application from someone that works with 2 machines
>(I forget who makes it or what it's called) where it constantly mirrors
>changes between machines and when the primary box fails the secondary
>changes it's IP address and reboots as the primary.
>
Well, that wouldn't exactly be transparent, you'd have this gap while it was
rebooting.

>clustering is used more for increasing the capacity where you need multiple
>machines just to handle the load. There is some sort of clustering setup
>where all the machines in a cluster can share exactly the same IP address
>but that's all I know about it.

OK, so following up the replication idea... Replication is where you have 2 or
more SQL Servers that all have exactly the same data all the time -- to keep
the example simple here, think of two, one is the publisher and one is the
subscriber, the publisher is the one the users are working in, and it
automatically replicates all changes to the subscriber all the time.  (There
are other flavors, e.g. you can have mobile users as the publishers and the
central box as the subscriber, or mobile users as the subscribers that PULL
from the publisher when they want, etc etc)  This eliminates the need for that
application you mentioned above, SQL Server is taking care of keeping them in
sync.  So let's say you have this, what's the most transparent way to switch to
the standby box if the primary blows up?

>
>Geo.
>

--- BBBS/NT v4.01 Flag-5
 * Origin: Barktopia BBS Site http://HarborWebs.com:8081 (1:379/45)