And I agree a 100 percent!
What makes that so interesting: I had the same discussion with a customer last week!
They plan to move to the cloud (Office 365) soon. However, the customers business needs the mail service available. Now!
Here is my approach for the customer:
- Install a second Exchange 2013 Server (virtual)
- Establish a HAProxy based Load Balancer
- Configure Database Availability Groups (DAG)
- Configure a Witness-Server (Existing HA solution was used) – Could be a cheap Azure Service as well
They already had a good and solid hypervisor in place and the cost of an additional server was nearly nothing.
They don’t had a Load Balancer, because services that might need them are already cloud based and they had the important services clustered.
A solution like this is relatively easy to build, and the migration is also not that hard to do. With only a few changes and a minor outage, they increased the availability of the Mail-Service dramatically. And everything was very cost effective. Using a Load Balancer also brings some more advantages. You don’t need to use DNS based High availability and you can increase the availability of other services to.
And in my opinion, the invest is well spent. With the next updates, the higher availability will pay out. There are no Downtimes during maintenance or while they apply patches, updates or cumulative update (CU).
And as you might have heard: Microsoft releases “Quarterly Exchange Updates“.
Even when they start the Office 365 Cloud movement, the new solution will be a benefit for the customer. While the Azure Cloud is highly available by definition, the hybrid solution might be the single point of failure if the service is based on a single Exchange Sever without Database Availability Groups (DAG).
And in this case, the newly established Load Balancer service makes it very easy and effective to bring higher availability to services they will use later. E.g. Active Directory Federation Services (AD FS).
So, if you still have a single Exchange Server and the question comes up: “How Small is Too Small for a Database Availability Group?“, the answer should be relatively easy to answer.
And if you already have a Load Balancer and/or a Virtualisation in place, the question is even easier to answer.
So do the Math:
- How available is your service today?
- How important is the Mail-Service for your business?
- How much does a downtime of let’s say two (2) hours might cost you?
If you ask me, there is no to small! As long as you need a reliable Mail-Service!