| Author |
Message |
SilverICE
Guest
|
 Posted:
Fri Mar 11, 2005 11:49 pm Post subject:
Design Review |
|
|
Thinking from an administration and recoverability standpoint, what is the
general opinion on whether to "sprawl" an Exchange organization over several
servers or to have "all of your eggs in one basket"? I'm trying to come up
with a recommendation and am struggling with this question myself.
Our company's Exchange policy is that they will be clustered, SAN attached
servers. We have about 2500 mailboxes and NO mailbox limits. That means we
have very large databases (~300 GB). My goal is to get the DBs down around a
more manageable size, say 100 GB.
Scenario 1 "sprawling": Exchange Server 2003 installed as two node,
active/passive cluster server. Server comprised of four storage groups, with
one mailbox store per Storage Group. We will need four servers. After
computing the hardware costs, this scenario will cost about $370,000.
Scenario 2 "all eggs, one basket": Exchange Server 2003 installed as two
node, active/passive cluster server. Server comprised of four storage groups
with four mailbox stores per Storage Group. We will need one server. After
computing the hardware costs, this scenario will cost about $230,000.
Scenario 3 "somewhere in the middle": Exchange Server 2003 installed as two
node, active/passive cluster server. Server comprised of four storage groups
with two mailbox stores per Storage Group. We will need two servers.
Estimated hardware costs = $300,000.
I'm leaning towards scenario 3 just because I think scenario 1 is wasteful
and scenario 3 could be dangerous if I had a catistrophic disaster to my
cluster. However, if money was no object how would you design it?
|
|
| Back to top |
|
 |
Al Mulnick
Guest
|
| Posted:
Sat Mar 12, 2005 1:10 am Post subject:
Re: Design Review |
|
|
Regardless of the money factor, I'd favor scenario 3 with 1 a close second.
The reason? Backup and restore times. The more condensed you are the harder
it is to put that back faster. A two-node cluster is likely enough (if
properly managed) to give a lot of availability but will cost more than two
separate standalone servers. Storage is a big concern as well, but it's the
recovery that has me the most concerned with the layout. Spreading that out
is how Exchange does best.
Al
"SilverICE" <SilverICE@discussions.microsoft.com> wrote in message
news:58D638CA-4323-40AD-86B9-69C90A62AD76@microsoft.com...
| Quote: | Thinking from an administration and recoverability standpoint, what is the
general opinion on whether to "sprawl" an Exchange organization over
several
servers or to have "all of your eggs in one basket"? I'm trying to come
up
with a recommendation and am struggling with this question myself.
Our company's Exchange policy is that they will be clustered, SAN attached
servers. We have about 2500 mailboxes and NO mailbox limits. That means
we
have very large databases (~300 GB). My goal is to get the DBs down
around a
more manageable size, say 100 GB.
Scenario 1 "sprawling": Exchange Server 2003 installed as two node,
active/passive cluster server. Server comprised of four storage groups,
with
one mailbox store per Storage Group. We will need four servers. After
computing the hardware costs, this scenario will cost about $370,000.
Scenario 2 "all eggs, one basket": Exchange Server 2003 installed as two
node, active/passive cluster server. Server comprised of four storage
groups
with four mailbox stores per Storage Group. We will need one server.
After
computing the hardware costs, this scenario will cost about $230,000.
Scenario 3 "somewhere in the middle": Exchange Server 2003 installed as
two
node, active/passive cluster server. Server comprised of four storage
groups
with two mailbox stores per Storage Group. We will need two servers.
Estimated hardware costs = $300,000.
I'm leaning towards scenario 3 just because I think scenario 1 is wasteful
and scenario 3 could be dangerous if I had a catistrophic disaster to my
cluster. However, if money was no object how would you design it? |
|
|
| Back to top |
|
|
SilverICE
Guest
|
| Posted:
Sat Mar 12, 2005 1:43 am Post subject:
Re: Design Review |
|
|
Good point with backup/restore times, and a point I did not include in my
original post due to its length.
We use snapshot backups and can currently backup 800 GB of data across 3
databases in about 10 minutes. Restore times would be similar. This time of
course would drastically increase if we had to go back and pull one off tape.
So, with that in mind, do you still feel the same way?
"Al Mulnick" wrote:
| Quote: | Regardless of the money factor, I'd favor scenario 3 with 1 a close second.
The reason? Backup and restore times. The more condensed you are the harder
it is to put that back faster. A two-node cluster is likely enough (if
properly managed) to give a lot of availability but will cost more than two
separate standalone servers. Storage is a big concern as well, but it's the
recovery that has me the most concerned with the layout. Spreading that out
is how Exchange does best.
Al
"SilverICE" <SilverICE@discussions.microsoft.com> wrote in message
news:58D638CA-4323-40AD-86B9-69C90A62AD76@microsoft.com...
Thinking from an administration and recoverability standpoint, what is the
general opinion on whether to "sprawl" an Exchange organization over
several
servers or to have "all of your eggs in one basket"? I'm trying to come
up
with a recommendation and am struggling with this question myself.
Our company's Exchange policy is that they will be clustered, SAN attached
servers. We have about 2500 mailboxes and NO mailbox limits. That means
we
have very large databases (~300 GB). My goal is to get the DBs down
around a
more manageable size, say 100 GB.
Scenario 1 "sprawling": Exchange Server 2003 installed as two node,
active/passive cluster server. Server comprised of four storage groups,
with
one mailbox store per Storage Group. We will need four servers. After
computing the hardware costs, this scenario will cost about $370,000.
Scenario 2 "all eggs, one basket": Exchange Server 2003 installed as two
node, active/passive cluster server. Server comprised of four storage
groups
with four mailbox stores per Storage Group. We will need one server.
After
computing the hardware costs, this scenario will cost about $230,000.
Scenario 3 "somewhere in the middle": Exchange Server 2003 installed as
two
node, active/passive cluster server. Server comprised of four storage
groups
with two mailbox stores per Storage Group. We will need two servers.
Estimated hardware costs = $300,000.
I'm leaning towards scenario 3 just because I think scenario 1 is wasteful
and scenario 3 could be dangerous if I had a catistrophic disaster to my
cluster. However, if money was no object how would you design it?
|
|
|
| Back to top |
|
|
Jim Schwartz
Guest
|
| Posted:
Sat Mar 12, 2005 7:10 am Post subject:
Re: Design Review |
|
|
Putting back the databases is one thing. Replaying those log files is
another. More users, more log files. If you're not going to set mailbox
limits, then 3 would be my choice as well. Those servers are going to grow
and you'll want a little extra overhead to handle and spikes in utilization
or DB growth.
"SilverICE" <SilverICE@discussions.microsoft.com> wrote in message
news:9FF3B1CF-4DE3-495C-8B38-40CE5FBCAC24@microsoft.com...
| Quote: | Good point with backup/restore times, and a point I did not include in my
original post due to its length.
We use snapshot backups and can currently backup 800 GB of data across 3
databases in about 10 minutes. Restore times would be similar. This time
of
course would drastically increase if we had to go back and pull one off
tape.
So, with that in mind, do you still feel the same way?
"Al Mulnick" wrote:
Regardless of the money factor, I'd favor scenario 3 with 1 a close
second.
The reason? Backup and restore times. The more condensed you are the
harder
it is to put that back faster. A two-node cluster is likely enough (if
properly managed) to give a lot of availability but will cost more than
two
separate standalone servers. Storage is a big concern as well, but it's
the
recovery that has me the most concerned with the layout. Spreading that
out
is how Exchange does best.
Al
"SilverICE" <SilverICE@discussions.microsoft.com> wrote in message
news:58D638CA-4323-40AD-86B9-69C90A62AD76@microsoft.com...
Thinking from an administration and recoverability standpoint, what is
the
general opinion on whether to "sprawl" an Exchange organization over
several
servers or to have "all of your eggs in one basket"? I'm trying to
come
up
with a recommendation and am struggling with this question myself.
Our company's Exchange policy is that they will be clustered, SAN
attached
servers. We have about 2500 mailboxes and NO mailbox limits. That
means
we
have very large databases (~300 GB). My goal is to get the DBs down
around a
more manageable size, say 100 GB.
Scenario 1 "sprawling": Exchange Server 2003 installed as two node,
active/passive cluster server. Server comprised of four storage
groups,
with
one mailbox store per Storage Group. We will need four servers. After
computing the hardware costs, this scenario will cost about $370,000.
Scenario 2 "all eggs, one basket": Exchange Server 2003 installed as
two
node, active/passive cluster server. Server comprised of four storage
groups
with four mailbox stores per Storage Group. We will need one server.
After
computing the hardware costs, this scenario will cost about $230,000.
Scenario 3 "somewhere in the middle": Exchange Server 2003 installed
as
two
node, active/passive cluster server. Server comprised of four storage
groups
with two mailbox stores per Storage Group. We will need two servers.
Estimated hardware costs = $300,000.
I'm leaning towards scenario 3 just because I think scenario 1 is
wasteful
and scenario 3 could be dangerous if I had a catistrophic disaster to
my
cluster. However, if money was no object how would you design it?
|
|
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in