Hello all, we would like to move to Exchange 2010. This is our currect scenario: 2 x CCR servers for db 2 x CAS/HUB in load balancing The databases are on local hard disks (of CCR machines) and all worked fine We have 70 mailboxes and the db size is around 65GB We would like to move to 2010 Can you suggest the best scenario that could fit our needs? I thought: 4 (2 x DAG, 2 x CAS/ HUB, using the cas virtual ip for balancing) or 6 servers (2 x DAG, 2 x CAS/HUB, 2 TMGs) We don't have a hw load balancer and we would not want to get it. Servers will be on the same place (same rack) Any help would be appreciated Thank you

Hello, looks like a good plan ;-) I'd prefer the TMG solution.... Greetings, Toni

Thanks Doni, also interested too hear more thoughts from other users with tech suggestions :)

Hello, The solution will depend on your requirements and what sort of high availability your require. A hardware load balancer is recommended - when you say CAS virtual IP, can you supply more information? Thanks

Hi ams this would be complete the scenario: Internet => Firewall (nat) => 2 x TMG (virt ip) => 2 x CAS/HUB (virt ip), 2 x DAG i would like to have the virtual ip of the TGMs natted internally, eg: Public ip 80.80.80.80 (on firewall) and TGM virtual ip (nat) 192.168.x.x So when the email is sent to the public ip on port 25, it ca reach the virtual TGM ip via nat i would also like to use the virtual ip of the CAS servers to have the high availability of the CAS/HUB (internally) What do you think? Thank you

Ok, do you mean the virtual IP on the firewall? If so, bear in mind you may want to load balance the RPC client access array for internal clients. Please see here for more information: http://technet.microsoft.com/en-us/library/ff625247.aspx Thanks

Hi ams, basically i want to load balance the TMG (with a virtual ip natted by the firewall) and the CAS array (with another virtual ip) The traffic enters the firewall and goes to the TGM virtual ip (natted with the private virtual ip of the TMG), then it goes to the CAS array (balanced as well) What do you think? Thanks

Hello, Yes from a high level that sounds good and if configured will give you HA for Outlook Anywhere/ActiveSync/OWA. You may like to look at the following whitepaper which gives more details about the scenarios publishing Exchange 2010 with TMG and UAG. http://www.microsoft.com/download/en/details.aspx?displaylang=en&id=8946 I note you state that you do not want a hardware load balancer. I would recommend looking at a load balancer in addition to the above if you have Outlook MAPI clients in your environment. Please see the following link which lists certified load balancers for Exchange 2010: http://technet.microsoft.com/en-us/exchange/gg176682.aspx Thanks

Hi Ams, thanks for your suggestions :) i have another question: from this thread: http://social.technet.microsoft.com/Forums/en-US/forefrontedgeiag/thread/a5e5a904-e5d6-4f35-87ac-12d96bda790b i noticed there is a way to use only 2 servers for high availability (DAG, CAS/HUB) and fully redundant. It seems should work. Can i use this scenario (with inly 2 servers for DAG cluster, CAS array and HUB) to achieve high availability and then add other 2 TMG servers to have HA on SMTP, OWA, EAS? In this way i would have 2 HA (for CAS and for TMG). Is it a feasible scenario ? thank you L

Hello The thread is not totally clear how this solution was achieved. It seems to be confusing OWA/EAS/OA and RPC client access array. Please see the below extract from technet (in particular the item in bold): Load balancing is recommended for high availability, failover, and for spreading the traffic load over multiple servers to help performance. When you choose a load balancing solution, consider the following: Windows Network Load Balancing isn't supported on Windows failover cluster servers. You can't use a Client Access array across multiple Active Directory sites. Instead, create two Client Access arrays and load balance separately within the sites. Hardware load balancers typically monitor return traffic, port availability, or service availability to ensure that servers that can't answer client requests aren't given network connections. Some load balancing solutions, such as ISA 2006 or TMG 2010, can't do RPC load balancing or monitor RPC services. These solutions aren't recommended unless all clients are connecting via Outlook Anywhere and all traffic is encapsulated inside HTTP. For more information about load balancing, see Understanding Load Balancing in Exchange 2010.

Ok thanks for your replies: from this document i see Microsoft support 2 servers scenario with a load balancer. http://technet.microsoft.com/en-us/library/dd979781.aspx#Two I would like to know if this Microsoft solution (with a hw load balancer) can give me HA on all services and if i can replace the hw load balancer with 2 TMGs (load balanced) Basically i need to have HA on SMTP, OWA and Active Sync (and of course the database) Thank you very much for your help :)

Yes that looks good and will give you HA for all services. You will need to use one of the approved h/w load balancers to achieve this and be supported. Note: with a 2 member DAG you will need a FSW which can be any other server in the environment to achieve quorum. You could add the 2 TMG's to this solution for OWA/EAS/OA. Internal Outlook clients will just use the H/W load balancer as the RPC endpoint. Create something like outlook.internal.domain namespace for these clients which will point to the VIP of the LB. Create an external namespace for OWA, EAS etc.

Ok thank you very much :)

Hi ams, sorry ... just a suggestion: What is the best scenario to have Exchange services in HA with a minimum hardware requirements? I mean: HA on database (and i can do it with DAG), CAS/HUB? And how i get HA on SMTP Inbound connections? Thank you :)

Hello Here is a link which shows a two node DAG (single datacenter) with HT and CAS co-located. http://technet.microsoft.com/en-us/library/dd979781.aspx#Two Note the load balancer is required in this design which will provide HA for RPC/OWA/OA/EAS availability etc. Inbound SMTP could be routed through the firewall to maybe TMG/ISA and then to the load balancer virtual IP. Note even with this design you are not designing for a site failure etc becuase the exchange environment would be in the same phsyical site and, you have just one load balancer which is a single point of failure. BUT it's all down to whats best for your environment and your SLA's. Thanks Tony