Exchange 2010 DAG Routing and CAS High Availability

In this blog post I wish to run through two key areas most Exchange Administrators do not know or have not come across with Exchange 2010. These are in my eyes, limitations in the product and need to be taken into close consideration when planning an Exchange 2010 enterprise infrastructure. As they are limitations to the product, it has not been documented very well by Microsoft and supporting Exchange communities.

Below we will cover:
- Database Availability Groups Routing
- CAS High Availability for Remote Sites

DAG Routing

For those of you who are not up to speed on Exchange 2010 Database Availability Groups I highly suggest you read my blog post on Exchange 2010 Database Mobility.

http://clintboessen.blogspot.com/2009/08/exchange-2010-database-mobility.html

As I illustrated previously, up to 16 servers can be a member of a DAG. Exchange mailbox servers have their own replication engine for replicating database transaction logs between servers running on TCP 64327 by default (can be changed). However there is no logic between this replication - take the following diagram for example.



Here is a DAG with only a single mailbox database. A passive copy of this database lies in every remote site. With DAG replication there is no smarts, there is no bridgehead server, no way of optimising which way replication traffic should go. The active mailbox database MDDB01 lies in London. This database is replicated everywhere. It needs to get to both Russia and Germany. Instead of copying it from London to Russia then Russia to Germany, it will copy it from London to Russia then London to Germany meaning the replication traffic needs to pass over the London to Russia path twice! This is something they may improve in the next version of exchange.

You can enable "encryption" and "compression" of replication traffic using powershell... the compression can help a little keeping bandwidth down.

CAS High Availability for Remote Sites

In Exchange 2010 your outlook clients talk to a Client Access Server for MAPI connectivity. The CAS then passes the requests on to the mailbox server. When you open outlook for the first time, Autodiscover will setup your outlook profile providing you with the closest CAS server to where your mailbox resides. These can be CAS Arrays, A bunch of CAS servers in a load balanced scenario.

What happens if you have a remote site with a single all in one exchange box running Mailbox, Hub Transport and Client Access roles that's a member of DAG replication for backup and off site recovery, and that server failed. Would users automatically fail over? The answer here is no. The reason being is because the DNS name they are pointing to is the local CAS server in that site, which in this instance is not highly available. The CAS server must be a member of a CAS Array to achieve automatic high availability. However saying this, the mailbox database is still online, as the next site in the priority list holding the passive copy has now become active. If the outlook clients changed their exchange server address to another CAS server in the organisation that is online, they will be able to get onto their email! Note that Outlook Anywhere/Outlook Web Access and Active Sync will always be highly available.