Re: Geographical Clustering
- From: "John Toner [MVP]" <jtoner@xxxxxxxxxxxxxxxxxxxxx>
- Date: Tue, 10 May 2005 18:06:18 -0400
As far as I know, you really only have 2 options for SAN storage in a
geographically dispersed cluster:
1) Use ISL between datacenters to stretch the fabric. Essentially, you still
have a single storage array at one location and stretch the connection to
the storage from your secondary site to the shared storage. In my opinion,
this is not much of a solution as you gain no HA benefit from this.
2) Use a storage vendor's synchronous mirroring solution, like EMC's SRDF.
By doing so, this also complicates things on the cluster side as you'll need
to find a way to control the synchronization between sites and coordinate
this with MSCS. EMC offers SRDF/CE (aka GeoSpan) to glue the two
technologies together.
The difference between an IP based replication and SAN based replication is
mostly reliability and consistency of your data. IP mirroring solutions are
not true synchronous replication technologies so you might have issues
getting things to work properly at your DR site should a site disaster
occur. SAN based replication technology can offer a true synchronous
connection between sites. Additionally, there might be "split brain" issues
with IP based mirroring if the software allows MSCS to take ownership of the
device should all IP links get cut between sites.
You could also throw in the standard arguments here for host-based mirroring
verses hardware-based mirroring. Additionally, SAN based replication is MUCH
more expensive. Typically, budgets will lean towards IP based replications
;-)
You should also consider supportability of your clustering solution. Your
geographically dispersed cluster solution should be qualified by MSFT and
posted to the Windows Catalog. Last time I checked, I did not see any IP
based solution there.
Regards,
John
"Jo Haydon" <johaydon11@xxxxxxxxxxx> wrote in message
news:%236rOEJYVFHA.580@xxxxxxxxxxxxxxxxxxxxxxx
> Hi,
>
> We're looking at implementing a geographically dispersed cluster and would
> really like to know what options we have for the storage back-end part of
> the setup. We'd like to have identical storage hardware in 2 different
> datacentres in London which both contain exact copies of the same data ie.
> they are mirrored. Previously we were looking at doing this using software
> such as NSISoftware's Double-Take but are now looking at maybe venturing
> into the SAN arena. I suppose my main question is really how a SAN setup
> would differ from just 2 servers running mirroring software over a
standard
> ethernet network? Each storage server would have multiple disks in a RAID
> configuration and we would like to know how this RAID is extended over to
> the other datacentre?
>
> Thanks - Jo/
>
>
.
- References:
- Geographical Clustering
- From: Jo Haydon
- Geographical Clustering
- Prev by Date: File Server Cluster
- Next by Date: Re: File Server Cluster
- Previous by thread: Geographical Clustering
- Next by thread: Issues with printer spooler and shares on W2K SP1 cluster...
- Index(es):
Relevant Pages
|
