Im looking to setup 2 exch2013 servers as active / passive DAG across AD sites - and implement HA/failover.

ServerA (active) in site 1 (primary site)

ServerB (passive) in site 2 (secondary site)

I want to set these up as stretch DAG. And have across site failover / resilience.

Questions:

First, will this design work? 400 user mailboxes. Is there a better design?

Which site should the FSW (file witness) be in?

How will other sites access serverB  if Site 1 goes offline. DNS round robin? I hear this is slow. Whats best practice for re-routing users to site2 ?? Options?

I've used Load balancers in 2010 but they say thats not needed in 2013. - true?  Because its active passive, I guess there is no CAS array?

Help?

(1) I have to ask you back , is there any user on site2 ?
If you want to active all DB in site 1. It's mean all user at site 2 have to access outlook in site 1.

(2) Witness Server , better you put this server with the active primary site.
please read this link. this link is most perfect explain about DAG design and witness server.
http://www.shudnow.net/2011/08/05/exchange-2010-site-resilient-dags-and-majority-node-set-clustering-part-1/

(3) If site 1 goes down like a disaster (Exchange and witness go down). You have to manual "switch over" for active database on site 2. This link also explain everthing about it.
http://www.shudnow.net/2011/08/05/exchange-2010-site-resilient-dags-and-majority-node-set-clustering-part-1/

(4) DNS round robin is using for CAS not for DAG. but i prefered "Hardware Loadbalance" if you have more than 1 CAS per site (windows NLB is okay , but for better choice you choose use hardware loadbalance")

(5) Like I said "Load balance" is not about DAG setting. Load balance is using for HA for CAS.
and you're right 2013 don't have casarray. But you have to make it clear. dont have casarray in 2013 is mean "In Exchange 2013 each database don't connect to CAS via cas array FQDN anymore".
So it's mean , Exchange 2013 still need to make HA for CAS with NLB.