We currently run Exchange 2007 on Server 2008. My boss is concerned that the Exchange server isn't fault tolerant and has asked me to look into costs for a better solution. Having done some research it seems that Exchange 2010 Database Availability Groups (DAGs) are the way to go. However, cost is an issue. We don't have a SAN, or any virtualization, so I'm just thinking about having two servers in the DAG. Our main concern is the amount of downtime in the event of a server failure.
Looking at documentation on the web it seems that this is okay, but will need a minimum of three servers; two in the DAG and one for the heartbeat. I've seen two possible setups; one where the servers in the DAG just run the mailbox role and the client access and hub transport roles are on a separate server. Also, I've seen a scenario where the mailbox, client access and hub transport are on both servers in the DAG and some form of network load balancer is used to direct client access (either ISA 2006, or some other hardware or software solution). With both of these scenarios there still seems to be a single point of failure (client access server or network load balancer).
I've had a look at HPs Exchange 2010 Sizer software, which can be used to get a costing for Exchange. Using their tool I'm getting a quote of £28k, which is way too much.
Has anyone setup Exchange 2010 DAGs? I know I need Server 2008 Enterprise for the OS, but Exchange 2010 standard edition will be fine for our needs (don't need more than 5 databases). I'm looking at DAS (we can't afford a SAN at the moment) and I'm after the most cost effective solution.
Many thanks if you've read this far. If you have, perhaps you could offer some advice? Any, and all, would be much appreciated.
We're looking at putting in a fully resilient 2010 solution in the very near future, but its not going to be cheap.
We'll have two DAG servers in each of our three main offices, with two load balanced CAS servers in each and one hub transport in each - making 15 in total (not counting the upgrade to Forefront threat management gateways over our exiting ISA's too). That way, as the DAGs replicate between offices, the CAS roles and transports all work together then there's no single point of failure.
Personally I haven't found a way round spending loads of money, so I'll be putting the costs forward and seeing what happens. If they want it they've got to pay for it.
Note: if you are gonna go vmware, use a fibre connection to the disk. iSCSI connections aren't supported when using load balancing between vms (as our boston office has just discovered...)
Thanks Domino. I fear this is going to involve a fair wedge of cash!
This depends on how many users you are supporting on the exchange box as many roles can be installed on one server and replicated to another server. The prev. post by Domino seems very interesting that you need 15 servers to accomplish a simple replication seem really strange. I would have thought that you could combine the roles into one server and only have the edge transport installed on a seperate server which will be responsible for your outgoing and incomming mails and then redirecting them to the right server.
We're looking at approx 1600 users with mailbox quotas of 200 MB each. I'm thinking the best option might be to have 2 mailbox servers in the DAG, with a single server as the client access and hub transport. I know this doesn't offer true fault tolerance, but I'm guessing the hardware requirement for the client access and hub transport server would be fairly low? Also, the data on this server would be quite small (in comparison to the mailbox servers) and so a disaster recovery would be easier?