Link to home
Start Free TrialLog in
Avatar of Sam Newbie
Sam Newbie

asked on

Exchange 2013 High Availability for Small Bussiness

Hi, Just wanted to see what others in similar scenarios are doing with similar setups.

We have a 2x Server 2012 R2 node Cluster with Hyper-v role with shared ISCSI storage supporting 40 users with around 60 mailboxes running the following VM's :

Active Directory, Exchange 2013, GFI Mail Archiver and File & Print.

Going Office 365 was not an option to consider for us at the time of implementation (nor is it now) for various reasons and On-Premise was the only option.

We understand Microsoft's Preferred Architecture and recommended best practice for Exchange 2013 High Availability inside a Virtual Machine requires a DAG and not to make the machine Highly Available at the Hyper-v cluster level.  However we see running the additional DAG member as an extra resource and an overkill for such a small setup.  We appreciate that High availability comes at a cost but just wanted to see what everyone else is doing and whether or not anyone else has steered away from the preferred architecture and made Exchange 2013 Highly Available at the Hyper-v cluster level without a DAG.

Our concerns are mainly around Database Integrity, etc.. during Live Migration.

Any help, advice or experience is greatly appreciated
Avatar of Will Szymkowski
Will Szymkowski
Flag of Canada image

Microsoft has these preferred architectures in place for a reason and they are best practices. Personally I would not go outside of the guidelines for architectural designs that Microsoft has provided.

Also as for HA using DAG, I really don't see the need for 60 mailboxes. You have to take into account the following when designing an architecture with DAG.
- Additional Server Licenses Cost
- Additional Resources for the second node (CPU, RAM etc)
- Additonal Storage for second node
- Added management/complexity (Exchange 2013 has definitely simplified the DAG process but it is still added configuration you need to deal with)
- Added Replicaiton Network (MS recommend using an isolated network for replication which could also had added cost associated to new network hardware to isolate IP Subnets)

Will.
We've been delivering SBS 2008 Standard, SBS 2011 Standard, Exchange 2007/2010/2013 Standard standalone, and other server roles in Hyper-V clusters since Hyper-V RTMd.

We have not had any issues with that setting other than once or twice where someone or something pulled the plug on the entire cluster.

That being said, one can run with two nodes and a DAS based SAS JBOD for a very inexpensive cost. This "Asymmetric Cluster" uses Storage Spaces to arbitrate storage between the nodes. Two nodes with a pair of SAS Cables connecting them to the JBOD have 96Gb of zero latency bandwidth at their disposal. This _kills_ iSCSI for performance plus DAS is dead simple to deploy and manage versus iSCSI's complexities.
Avatar of Sam Newbie
Sam Newbie

ASKER

Thanks Will and Philip,

So other than the DB having a "Dirty Shutdown State" what prevents me (Other than the microsoft Preferred Architecture) from Live Migrating an Exchange 2013 VM and making it HA via the Hyper-V Cluster?
ASKER CERTIFIED SOLUTION
Avatar of Philip Elder
Philip Elder
Flag of Canada image

Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
Perfect, just what I thought. Do you think while we grow in size and as our mailboxes increase in numbers there is a threshold where we would need to implement a DAG? We'd obviously have to ensure we have the appropriate licensing, resources and storage to support it.
Thanks
I know I am a little late to the party
 
DAG's are for high availability.  The power of a DAG, is when MBX server 1 fails, all the active Mailbox databases move to MBX server 2, without users noticing, also MBX databases can also be stored locally, SAN, DAC etc, servers could even have DB's stored in different locations as long as the partnership is created.  The number of databases, or database size is not as relevant to a DAG, as long as you have adequate space where ever you are storing the mailbox database and have enough IOPS to support the number of connections.  Also the decision to implement a DAG usually comes from business requirements.  You could have 1 mailbox or 2000 mailboxes if the business deems that the system with 1 mailbox must not failand have no service disrutpion then a DAG is needed, to support just 1 mailbox.  Performance is based on your disks subsystems.
Not late at all, really appreciate your reply Jessie.
For the meantime we will try to do without a DAG till required.
thanks again
Amazingly quick reply!
Sam,

I had a follow-up on the sizing question all ready to go at the shop but got called out. TechNet and TechNet Blogs have a number of great performance and sizing posts that will help.

Exchange 2013 is designed to run in RAM with hundreds of mailboxes on a small spindle drive set. So, the larger your org grows the more RAM the Exchange VM would need. IOPS would become a bit of a question at some point down the road but probably not before the end of the solution's life.
Thanks Philip, we will implement a DAG in either case and scale the VM's as we grow.  There's an amazing amount of information out there for Exchange starting from the Calculator.
Sam,

An option that one can implement on 2012 R2 is a guest cluster. There are ways to go about it with Exchange to allow for independent storage configurations too.

And yes, info is out there but sometimes putting it together into a coherent package can be challenging. ;)
Hi Philip, agreed.  Can get a little confusing especially with conflicting information and very little time.

What would be the advantages of a guest cluster in comparison to either a HA Exchange VM or a DAG?
A guest cluster setup means running a cluster within the Hyper-V cluster. So, your two Exchange nodes connected to independent CSV based storage for the DAGs. It would bear more research though.