SQL 2000 Replication Best Practices

[Elroyskimms]

I have a production DB server that I want to mirror/replicate to a
second server to be used as a fail over in case of a localized network
outage. I'm in the beginning of the planning stages and have no
experience in SQL replication. Where can I find some "Best Practices"
for doing this? During replication, is the data encrypted in any way or
do I need to secure the link myself (VPN or a P2P connection)? Details
below...

MS SB Server 2003
Hosting various websites that link to local database.

Would like to set up a backup system, running MS SB Server 2003,
hosting those same websites. I can replicate the websites using NTFRS,
that's not a problem. If the primary system goes down, I want the local
websites on the backup system to use the local dbase. Is that the best
way to do it, or should I configure it some other way? The goal is
redundancy, not load balancing.

Thanks in advance for your suggestions!

(BTW - I searched this group for "Best Practices" but couldn't find
anything related to this type of scenario)

[Geoff N. Hiten]

Replication copies the DATA but does not copy the entire DATABASE.
Generally, a replicated copy is not suitable as a transactional data store
for DR purposes. If you insist on using Replication, you might want to read
this book: http://www.nwsu.com/0974973602.html. It goes into detail on how
and why you whould want to use any particular replication feature.

Geoff N. Hiten
Microsoft SQL Server MVP

"Elroyskimms" <elroy...@yahoo.com> wrote in message
news:1118338333....@z14g2000cwz.googlegroups.com...

[Aaron Asbra]

Geoff,

Thanks for the reply. If replication is not a suitable method for
creating a roll-over for disaster recovery, what method would you
recommend? We're just getting started on the project, so I'd rather do
it right the first time.

-E

*** Sent via Developersdex http://www.developersdex.com ***

[Geoff N. Hiten]

How much money do you have?

Seriously, the more you reduce the impact, the more expensive it gets. Log
shipping to a warm standby server can take you from several hours downtime
to maybe one or two hours and isn't too expensive. Using a geographically
separate SAN mirroring solution can make the failover automatic and very
brief but it will cost a lot. Determine the actual cost of downtime and the
risk of downtime. Then figure out how much it is worth to mitigate that
risk by reducing the detection and correction portions of an outage. Plan
and purchase accordingly. The real key is not having the "best" Disaster
Recovery plan in the world, it is having a plan where management and the
technical people agree on the expectations and the costs.

Geoff N. Hiten
Microsoft SQL Server MVP

"Aaron Asbra" <elroy...@yahoo.com> wrote in message
news:%23ktr6ld...@tk2msftngp13.phx.gbl...

[Aaron Asbra]

Geoff,

Thanks again for the info. Lets just say that we had a 5 our outage the
other day and the money lost would have been enough to cover a brand new
quad Xeon, U360 SCSI RAID server with all the bells and whistles and
still have money to spare. I'm sure that's chump change to some
enterprise solutions out there, but this is a small shop and the system
literally is the business. If it is down, we are not only not making
money, we are also losing money.

I'm not familiar with SAN mirroring, could you direct me to some
material that would be usefull? As far as the geography, our primary is
in California and the backup will be housed in North Carolina.

A couple of hours downtime is not acceptable for this application. We
really need something that will handle the failover automatically.
Again, thanks for your help!

-A

[Geoff N. Hiten]

Nothing is chump change if it is critical to your business. It sounds like
a basic SQL cluster should be the starting point along with log shipping to
a remote site. This is a fairly normal configuration for a small to medium
sized shop. The cluster handles most of the immediate failover problems
automatically and if you have a total site disaster, you manually shift to
your backup location. I would avoid SCSI based solutions and go with a
fiber-channel SAN for a lot of reasons. Dell and HP have "canned" cluster
designs that are good for small and medium enterprises. I would definitely
start with your favorite major vendor and work with them. If you want some
type of geographic real-time recovery, make sure your vendor knows that in
advance so it can be added as a future step. I wouldn''t try and do it
right off the bat. Build your cluster and get used to what it can do for
you, then extend and expand it.

Geoff N. Hiten
Microsoft SQL Server MVP

"Aaron Asbra" <elroy...@yahoo.com> wrote in message

news:%2329x6ye...@TK2MSFTNGP09.phx.gbl...