Re: Storage Area Network
- From: "Al Mulnick" <amulnick_No_SPAM@xxxxxxxxxxx>
- Date: Sun, 26 Jun 2005 11:31:02 -0400
Exchange 2003 running on Windows Server 2003 (currently backed up using
exmerge script to .pst files)
80GB Mail Store
IBM X225 Server
EMC Clariion SAN (I can't remember the exact model, but I know it's one of
their lower-end units)
Our company has around 95 employees
Backup to PST? Why?
Do you expect less or more space to be taken if you take a backup via
Exchange-aware backups? You did remember to put Exchange logging into
circular mode for this correct? If not, you may want to check out the logs.
<G>
Backup to disk is a great option for many situations. There isn't enough
information here to understand your requirements or what type of
recommendation would be better. However, it's common practice to backup the
application with the supplied backup software. The media it writes to is
more important to you than it is to the application and is part of the
equation that gets factored in when you decide how long it will take to put
it all back should one of your disaster scenarios occur.
The Cariion SAN system is fine and is a low(ish) cost system. If it's
already being tasked by other applications, I too would question what you're
trying to do. Both Oracle and Exchange are highly disk dependent for
performance. Exchange is likely more dependent due to the application usage
characteristics (semi-random read/write vs. known usage pattern in a
SQL-type of db system), IMHO. Exchange is disk cache unfriendly in many of
its operations.
For both Oracle and Exchange, you'll realize performance gains by
concentrating on disk performance vs. disk space. That's not how SAN's are
sold I know, but it is how to make them work. There's only so much buffering
that can be done before you must go to disk for the data or to write the
data. Read-ahead and behind and write caching are not very effective with a
semi-random db technology.
As for the idea of geographical dispersed clustering I would argue that cost
is going to be much higher than adding spindles and switches to the current
environment. You have the added burden of ensuring that the data is
replicated real-time in that kind of setup (as you currently describe it).
For most of my customers I usually suggest a different approach. I typically
suggest figuring out what is important to them, assign the value of that
information and try to distill it to minutes if possible. This gives an
idea of how far they need to go with such a solution. I've often found that
getting back to a working state is far more important than getting the data
back right away and given the cost diferential between the two, I have
rarely seen where it's important to get both back that fast across distance.
Local datacenter clusters are a different story, but trying to do this
across the WAN is expensive and complicated in most situations.
For many, the 'dial-tone' (I hate that moniker; I've made many systems more
reliable than the phones in the same companies) scenario is what they're
really after, where a datacenter fails and the bring the service back up in
about 15-30 minutes. They're users would of course have to have
desktops/laptops and access to the network, but that's another story. They
then implement a recovery via tape or disk, to trickle back in the data to
the last point of a backup. The effect is that users are functional, but
without legacy data for the amount of time it takes to restore the tapes to
the mailboxes. Very few companies need the data real-time and can afford to
have the duplicate systems, hardware for the users, access to the network,
etc.
Hopefully this is helpful. If not, please let me know how I can be more
helpful.
Al
"Branden Bell" <bbell_97@xxxxxxxxxxx> wrote in message
news:ecc1wiZeFHA.1288@xxxxxxxxxxxxxxxxxxxxxxx
>A couple of my co-workers are intent on using our low-end EMC SAN as a
>backup to disk solution for our Exchange data.
>
> The concern is that the SAN was not purchased with the idea of ever using
> it for Exchange backups. The current Oracle data on the SAN is already
> causing performance issues that have not been resolved. The cost
> associated with the purchase of an extra host bus adapter and fiber
> channel switch is also a concern since our long term plan is to implement
> software and hardware that will automatically replicate Exchange data
> between two sites and be setup to automatically failover from primary to
> secondary.
>
> I would really appreciate it if some of you can respond with your
> feedback. Here are the details of our configuration:
>
> Exchange 2003 running on Windows Server 2003 (currently backed up using
> exmerge script to .pst files)
> 80GB Mail Store
> IBM X225 Server
> EMC Clariion SAN (I can't remember the exact model, but I know it's one of
> their lower-end units)
>
> Our company has around 95 employees.
>
>
>
>
>
>
.
- References:
- Storage Area Network
- From: Branden Bell
- Storage Area Network
- Prev by Date: Re: Cannot connect to remote host after upgrade to Exchange 2003
- Next by Date: RPC over HTTP- multiple logons
- Previous by thread: Storage Area Network
- Next by thread: RPC over HTTP- multiple logons
- Index(es):
Relevant Pages
|
