MEC Day 3

After 3 ½ days it’s all over. As I write this I’m sitting in Orlando International Airport trying to lose a couple of hours before my flight boards!

Todays interactive sessions have been varied and interesting, here is a brief overview:-

Exchange 2013 Load Balancing

This is a topic that comes up regularly when I am deploying Exchange 2010. The general reaction when discussing LB is that it’s either Expensive or unreliable.

Exchange 2010 had no active concept of an array, it was essentially just multiple, individual servers with no awareness of each other, this has changed in 2013.

The Exchange 2013 CAS role is essentially an intelligent proxy, all the client rendering technology is now located in the Mailbox role, meaning that the CAS servers could, in theory, be load balanced by Round Robin DNS or Windows Network Load Balancing with no problem, until there is a service failure. This is where true “Service Aware” Load Balancers come in, with WNLB or RRDNS the request would still be sent to a server that was responding to TCP requests but Exchange services had failed. Intelligent load balancing would detect the service failure and not pass any traffic to that node until the issue is resolved.

Another challenge with WNLB is that if using client affinity, it will see an entire remote IP subnet as a single client, routing all traffic from that subnet to a single CAS server – not an optimal solution in a large estate but still valid for small farms (e.g. 4 servers).

The use of Layer 7 load balancers is no longer necessary but is still supported.

The 2013 CAS role, when used in a co-existence environment with 2010 will perform service monitoring of all discovered 2010 CAS servers every 60 seconds to enable it to effectively proxy traffic to the most appropriate host.

Another discussion broke out around TMG and options moving forward, but nothing more came out of this then from the day 2 discussions other than that SSL offloading is not supported in 2013 RTM.

Public Folder Migration

Much has been made of the “Modern Public folders” in Exchange 2013, especially as 5 years ago Microsoft stated that Public Folders wouldn’t be in the next version of Exchange (2010).

From a client perspective, the Modern Public Folders will appear and behave exactly as the old ones they currently use. Any LoB applications using Public Folders should still work, although some MAPI calls have changed slightly.

From an Admin perspective, things have changed.

Modern Public Folders are now stored in Mailboxes and, as such, can be protected by a Database Availability Group.

As they are stored in a Mailbox Database the old Multi-Master replication model has gone. Only one copy of a database can be active at any time (standard DAG behaviour), so this might introduce performance challenges for some clients.

Migration from 2010 to 2013 needs to be performed as a Cutover – it is not supported to have old & Modern Public Folders co-existing.

A number of PowerShell scripts are included to ease the migration to Modern Public Folders to ensure that all attributes, permissions, etc get migrated to the new hierarchy.

Ironically, the permissions model is “the same as before”!

Exchange 2013 Site resiliency

As this was another interactive session it didn’t cover many of the planning aspects it promised but being run by Greg Theil & Scott Scnoll it was a good session to attend anyway.

A lot of discussion around Datacentre RTO, such as “Does the clock start ticking before or after the operator gets called?”, “It’s far easier to have success with a 3 hour RTO that a 1 Hour RTO”

Greg was very persistent that until you have done multiple test failovers you cannot be confident that your failover plan works – ideally test monthly or at least quarterly.

Discussions around DAC mode (datacentre activation coordination) highlighted that when performing recovery the primary DC servers must be shut down to avoid the risk of “Split Brain”

Finally some product detail around DAG’s in 2013:-

  • The setup will automatically setup the Networks
  • Multiple subnets will be consolidated automatically
  • Additional NIC’s (e.g. iSCSI) will be shown as cluster networks and will need to be manually removed
  • The “Enable manual control” option allows the naming of networks & enable/disable replication etc
  • DAG member Exchange versions cannot be mixed – create a new 2013 DAG & use mailbox move.
  • There is no database portability between versions

Virtualisation in Exchange 2013

This was a little bit dry – not the best session to end on!

All two roles (!) in 2013 are fully supported in a virtual environment

Availability is built into Exchange (with DAG’s) so clustering of the virtual platform is just a mechanism for dealing with hardware failure and bringing servers back into service again.

Live Migration is supported, Quick Migration requires the server to be shut down before & cold booted after the migration.

Virtual machine Snapshots are not supported by Exchange (including Hyper-V replica)

Windows Server 2012 adds new functionality:-

  • Removed the 4 CPU limit – a problem in larger Exchange implementations
  • Allows the use of SMB 3.0 storage for Hyper-V, but still not directly for Exchange
  • Deduplication – not supported by Exchange
  • ODX (using intelligent SAN storage to move data natively) – not supported by Exchange (wouldn’t want multiple DAG nodes using the same storage/controller anyway)

That’s it for now. Once I get back to the UK and gather my thoughts and notes I may add some further detail.

Thanks for reading!


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s