Exchange 2007 - CCR or SCC cluster

Strunk · Joined: 07 Jan 2008 Posts: 4

Hello,

We are going to setup a new Exchange 2007 installation where
high availability is an issue. Right now we are looking at
CCR and SCC as one of the cluster models to use.

Here some input:

-1500 Users
-800 GB maildata
-HP EVA 6100 SAN with 56*300 GB discs.
-VmWare ESX 3 running on a HP Bladecenter C7000
with 5 x BL465c blades with 16 GB each.

Before we go any further we know that Exchange 2007 is NOT supported
on VmWare but we have decieded to go ahead anyway so if you leave
that issue out when you (hopefully) answer our questions.

We want to deploy the following:

- Two Clients Access servers running with NLB
- Two Hub Transport servers
- (no Edge servers as we have another solution for this role)

And then there is the Mailbox servers. We can't seem to make up our minds
about which cluster model to use CCR or SCC.
We are leaning a little towards the CCR as we here get copy of the database
if something goes wrong. You also have the possiblity to use the passive
database for backup.

Our 2 questions:

- If we want High availability and have the above equipment at our hands
what would be the best cluster model CCR or SCC ?

- We are also thinking about setting up a second data center and with that
in mind is it possible to "convert" a CCR or SCC to a SCR cluster ?

Any input will be appreciated,
Strunk

Archived from group: microsoft>public>exchange>setup

Oliver Moazzezi [MVP] · Joined: 02 Oct 2007 Posts: 583

I'd be interested to see how a 1500 user VM cluster runs, especially if you
don't have a 1 to 1 contention ratio (1VM:1Physical).

You should build a test lab using the production kit, and get testing. Then
get your superiors to sign off approval of a 1500 user VM SCC/CCR cluster
before sticking this straight into production. If it was me I would probably
keep the mailbox role physical, and VM the CAS and HT roles as you appear to
have more than enough blades - however thorough testing is your answer here.

Oliver

Scott Schnoll [MSFT] · Joined: 05 Aug 2007 Posts: 89

Response down below...

wrote in message @b29g2000hsa.googlegroups.com...
On Feb 1, 5:09 am, "Scott Schnoll [MSFT]"
wrote:
> CCR is preferred over SCC for many reasons, some of which you already
> pointed out:
>
> - CCR is less expensive to deploy than SCC
> - CCR protects your databases by maintaining a copy of each database on a
> separate server; SCC does not provide any protection for your databases
> - CCR provides service and data availability; SCC only provides service
> availability
> - CCR is less complicated to deploy; SCC is more complicated because it
> involves shared storage and more cluster knowledge
> - CCR reduces backup TCO by allowing you to offload backups from the
> active
> node to the passive node; SCC backups are possible from the active node
> only.
>
> Seehttp://technet.microsoft.com/en-us/library/aa997928(EXCHG.80).aspxfor
> some more information.
>
> - Show quoted text -
>CCR is less expensive to deploy than SCC? Really, idle physical
>servers (passive nodes) and multiple storage per cluster, how is that
>cheaper?

Yes, CCR is less expensive than SCC. Here's why:

1. CCR only requires hardware on the server HCL (what we now call the
Windows Server Catalog). This is true, even for geographically-dispersed
CCR environments. You do not need to purchase specialized hardware that is
listed in either the Cluster Solutions category or the
Geographically-Dispersed Cluster Solutions category. With SCC, you do need
to buy hardware from one of these categories (depending on your
configuration) and these are typically more expensive solutions than regular
standalone servers.

2. The passive node in CCR is not idle; in fact, it is the reverse; it is
idle in SCC and used extensively in CCR. With SCC you're required to at
least one passive node in each cluster. In SCC, the passive node is an idle
node. It is not doing anything other than standing by for when you need to
do maintenance on the active node, or when there is a sufficient failure on
the active node to cause a failover. In CCR, the passive node is absolutely
not idle. It is doing work for you, and very valuable work at that. It is
maintaining a second copy of each storage group. In SCC, you only have one
copy of your data; thus, if something happens to that copy, you have a
potentially lengthy and painful recovery process that can results in a lot
of downtime for your users. In CCR, you have a second copy being maintained
online for you. In the event of failure of the active copy, you have a
passive copy you can quickly activate, resulting in very minimal downtime
for your users. Moreover, CCR enables you to offload Exchange-aware
VSS-based backups from the active node to the passive node. This provides
you with a nice reduction in backup TCO, and gives you lots of flexibility
with your backup window. In SCC, you cannot backup your data from the
passive node because there is no data on the passive node. Thus, you're
always taking backups off the active node, which can interfere with
activities such as user operations, online maintenance, and messaging
records management.

3. With a 64-bit architecture, and many many JET/ESE optimzations in both
Exchange 2007 RTM and SP1 mean a substantial reduction in disk IOPS. As a
result, you no longer need a SAN for heavyily used mailbox servers.
Instead, you can use direct attached storage, which is much cheaper than a
SAN. Yes, you will be buying twice the storage as you pointed out, but (1)
you are buying less expensive storage, and (2) you are getting tremendous
value out of this extra storage because the extra storage is used to
maintain a second copy of your data online. That second copy means much
less downtime, which saves a lot of money in lost productivity, operations
costs, etc. That by itself should be worth the price of admission. Smile

4. SCC requires shared storage; CCR does not. As a result, CCR gives you
options for less expensive storage. You don't need a SAN with CCR; DAS
works great for CCR.

That said, you really want to look at your RTO and RPO for specific failures
to find out what solution is right for you. You also need to consider other
things that drive HA, such as support for large mailboxes.

Hope this helps.
--
Regards,

Scott Schnoll
Microsoft Corporation
This posting is provided "AS IS" with no warranties, and confers no
rights. Please do not send email directly to this alias. This alias is for
newsgroup purposes only.

Exchange 2007 SP1 upgrade CCR cluster Hello, I'm upgrading E2K7 RTM to SP1 in my LAB and I ran into a problem upgrading second node of the cluster. I have upgraded my other server (HT,CA) with no problem. I followed MS instructions for cluster with CCR upgrade and passive node was upgraded wi

Upgrading from an Exchange 2003 cluster to Exchange 2007 We currently have Exchange 2003 Enterprise running on a Windows 2003 (x32) cluster. I would like to know what's the best way to upgrade this cluster to Exchange 2007... We have purchased Windows Server 2003 Ent. (x64), and Exchange 2007 Ent. licenses, and

Exchange 2003 Cluster Setup (3A, 1P) Hello, I am about to install Exchange 2003 on the first physical node. Windows cluster is working fine with Quorum and three drive letters for mount points seen by all nodes (if it owns that resource of course). Question - does the first node need to own

Setting-up Cluster Exchange 2003 DR Server We have an existing pair of Exchange 2003 cluster at the primary site and we wanted to set-up the same Exchange 2003 cluster at a DR site. In order to avoid any downtime for the installation of Exchange Virtual Server (EVS) at the DR site, we would like t

SCR Cluster and Hub Transport Role Hello Is it possible to run hub transport, client access and the mailbox role on one server box with SCR cluster enabled? I know that I can't do this with CCR. Thanks for any help Sandro