Two server DAG solution with HLB

[ShackDaddy]

I'm putting together a two-server solution for a 200-user client with multiple geographic locations, and I want to implement DAG HA. I've purchased Windows 2008 Enterprise and Exchange 2010 Standard for both servers, and I'm currently planning on buying a Kemp LM2000. But I haven't really been able to find much information on comparative HLB's for E2010 use. Does anyone have any non-theoretical experience with these?

As I understand it, each site will have its own HLB and it will have an internal IP address. The firewalls will route SSL traffic to it and the clients will point at it for MAPI, and it will distribute that traffic to either of the CAS/Hub/MBX servers in the organization, based on prioritization by subnet and availability. That's the plan anyway, I just need to start getting to know the HLB's more intimately.

If anyone has advice or feedback, I'd appreciate it, since this is going to be an important scenario for a lot of mid-sized businesses and I want to be prepared.

Dave Shackelford
ThirdTier.net

 

[ntinlin]

Hi Dave,

You are looking at load-balancing across AD sites and subnets with only one server on each site?

Thanks
Neill

 

[ShackDaddy]

Not anymore. I've come to my senses. LB is only for multiple servers in a site, unless you want to make things overly complicated. Now I'm dropping the HLB devices and just stretching a two-database DAG across two sites. One server per site.

But I'm still very open to guidance.

Dave Shackelford
ThirdTier.net

 

[ntinlin]

I was wondering.

Only problem I can see (and I'm not an expert on 2010 yet) is that when you install a database it points itself to a particular CAS, and updates AD to that effect.

Thus part of the need for a CAS array for HA within a site.
Indeed once you set up the CASArray you need to update the databases to point to it instead of individual CAS.

What is unclear to me (and I haven't tested yet) is what happens if live DB moves to another site. Logically DB should still be pointing to other CAS within AD but I wonder if it is clever enough that if there is a CAS array defined for the site that it will change DB entry to use that instead. Would mean defining CAS array AD entry against your single CAS in each site I guess.
If that didn't work you'd probably need to update the DB CAS entry each time the database was flipped between the sites.

What we are going to be doing here is 2x mailbox servers in 1 AD site with separate load-balanced CAS+HT and a 3rd mailbox server with all 3 roles in DR. DR would not be set to come up automatically to avoid split-brain if WAN was lost so we'd do a datacenter switchover if we needed to go bring up that server.

Neill