Good Exchange infrastructure design
Evening
Just after some clarification on a few points
I look after a Company, around 200 users and growing and they currently have a single Exchange installation (2010, Typical) and they are looking to expand and implement HA for the Exchange. I probably have at least one physical server I can get out of the budget for now so here is the site layout and my ideas
Idea 1
Site A - Main office - 150 users
Install Physical server and virtualise with VMware
VM1 - CAS and HUB on W2k8 standard R2
VM2 - MBX on Windows 2k8 Enterprise R2
Existing VM (On different Physical host) that is running the Typical Exchange to be a CAS on standard W2k8 standard R2, with NLB enabled on the CAS/HUB (This should be fine as the NLB and WCS are on seperate servers)
Site B - Site office - 50 Users
VM 1 - Typical Exchange on W2K8 Enterprise
Stretched DAG between sites, the offline copy to be kept at Site 2
Idea 2
Site A
Two Physical servers
Physical Server 1
VM 1 - CAS/HUB (W2K8 standard R2)
VM 2 - MBX (W2K8 EntR2)
Physical Server 2
VM 1 - CAS/HUB (W2K8 standard R2)
VM 2 - MBX (W2K8 EntR2)
Load balanced CAS
DAG between MBX
Site B - Site office - 50 Users
VM 1 - Typical Exchange on W2K8 Enterprise
Stretched DAG between sites, the offline copy to be kept at Site 2
The second is if I can get away with the extra physical host, but does this look like a sound solution?.....
Any thoughts appreciated
Just after some clarification on a few points
I look after a Company, around 200 users and growing and they currently have a single Exchange installation (2010, Typical) and they are looking to expand and implement HA for the Exchange. I probably have at least one physical server I can get out of the budget for now so here is the site layout and my ideas
Idea 1
Site A - Main office - 150 users
Install Physical server and virtualise with VMware
VM1 - CAS and HUB on W2k8 standard R2
VM2 - MBX on Windows 2k8 Enterprise R2
Existing VM (On different Physical host) that is running the Typical Exchange to be a CAS on standard W2k8 standard R2, with NLB enabled on the CAS/HUB (This should be fine as the NLB and WCS are on seperate servers)
Site B - Site office - 50 Users
VM 1 - Typical Exchange on W2K8 Enterprise
Stretched DAG between sites, the offline copy to be kept at Site 2
Idea 2
Site A
Two Physical servers
Physical Server 1
VM 1 - CAS/HUB (W2K8 standard R2)
VM 2 - MBX (W2K8 EntR2)
Physical Server 2
VM 1 - CAS/HUB (W2K8 standard R2)
VM 2 - MBX (W2K8 EntR2)
Load balanced CAS
DAG between MBX
Site B - Site office - 50 Users
VM 1 - Typical Exchange on W2K8 Enterprise
Stretched DAG between sites, the offline copy to be kept at Site 2
The second is if I can get away with the extra physical host, but does this look like a sound solution?.....
Any thoughts appreciated
ASKER
Its for the offsite element of Exchange really if all of Site A went down then for a short amount of time B could come and staff could connect. An offline copy of the DAG sould exist here....
Site B would be the secondary MX location.
There is a DC at Site B, and this site could grow to over a hundred users so this would could then be broadened out to be like Site A...
Site B would be the secondary MX location.
There is a DC at Site B, and this site could grow to over a hundred users so this would could then be broadened out to be like Site A...
ASKER
Oh yeah, there is an MPLS between all of the sites - Fiber with EFM backup so connection speed is all good as is intersite connectivity...
First of all, I recommend you read the excellent articles on msexchange.org here:
http://www.msexchange.org/articles_tutorials/exchange-server-2010/high-availability-recovery/designing-site-resilient-exchange-2010-solution-part1.html
The design you suggest is in overall ok, but you you need to consider CAS arrays and maybe a HW loadbalancer as well.
But since you not have a lot of users, I think that you are making yourself a lot of extra work, and days of testing the final solution. If you have a limited budget (and time), I would consider to focus more on the underlaying infrastructure, in your case VMWare (vSphere with vCenter I hope). There is many things you can do to keep Exchange services more available that does not require change in the Exchange design. Most important is disk and network. And since Exchange is pretty useless without internet, you should focus to have redundant internet connection, that can failover automatically. That requires a big effort, depending on how you do it. If you want automatically failover between 2 ISP's and you still want your Exchange and other public servers alive, then you need to setup BGP on your firewalls and redesign your IP scheme. But in return, you get an automatic failover solution and no admin work is needed. And redundant internet connection is not only useful for Exchange, but for other services and for your users as well :) You could make a more simpler and manual failover between 2 ISP's, but that often requires you to change public DNS records as well, meaning there will be some downtime.
Storage and storage systems needs to be highly available as well, if you not already have a SAN with duplicate fabric in place already.
Lastly, you should setup a backup MX pointer to a service provider outsite of your sites, in case you will lose internet connection completely or your Exchange servers are down for a long time. In that case you will not loose any incoming mail when your services are up and running again.
http://www.msexchange.org/articles_tutorials/exchange-server-2010/high-availability-recovery/designing-site-resilient-exchange-2010-solution-part1.html
The design you suggest is in overall ok, but you you need to consider CAS arrays and maybe a HW loadbalancer as well.
But since you not have a lot of users, I think that you are making yourself a lot of extra work, and days of testing the final solution. If you have a limited budget (and time), I would consider to focus more on the underlaying infrastructure, in your case VMWare (vSphere with vCenter I hope). There is many things you can do to keep Exchange services more available that does not require change in the Exchange design. Most important is disk and network. And since Exchange is pretty useless without internet, you should focus to have redundant internet connection, that can failover automatically. That requires a big effort, depending on how you do it. If you want automatically failover between 2 ISP's and you still want your Exchange and other public servers alive, then you need to setup BGP on your firewalls and redesign your IP scheme. But in return, you get an automatic failover solution and no admin work is needed. And redundant internet connection is not only useful for Exchange, but for other services and for your users as well :) You could make a more simpler and manual failover between 2 ISP's, but that often requires you to change public DNS records as well, meaning there will be some downtime.
Storage and storage systems needs to be highly available as well, if you not already have a SAN with duplicate fabric in place already.
Lastly, you should setup a backup MX pointer to a service provider outsite of your sites, in case you will lose internet connection completely or your Exchange servers are down for a long time. In that case you will not loose any incoming mail when your services are up and running again.
ASKER
Thanks for your response, as I posted I am not concerned with the connectiosn as they have MPLS with internet breakout, with two connections to the MPLS and DSL lines onsite as well if needed
Deffo VMWARE Vsphere 5 and Vcentre is being pushed although they are yet to sign it off, but they don't have a massive VM environment yet.
Is my Exchange setup overkill then? Should I stay with Typical Exchange setups? I always believed seperating the roles is the best thing to do if possible?.....
Deffo VMWARE Vsphere 5 and Vcentre is being pushed although they are yet to sign it off, but they don't have a massive VM environment yet.
Is my Exchange setup overkill then? Should I stay with Typical Exchange setups? I always believed seperating the roles is the best thing to do if possible?.....
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
After a bit more planning I am going with the following (...if anyone cares!)
Site A
2x Physical servers
Each with 1200GB of storage (total after RAID 10)
VM on each server with 50GB for OS, 50GB for queues (CAS/HUB role) these will then be load balanced with NLB
VM on each server with 50GB for OS, 100GB for logs, 800GB for EDB's (MBX role) this leaves an additional on each of 150GB if needed.
My plan is to split the mailbox stores into 4 on each mailbox server for load balancing, with enough space on each to hold the passive database copies if they ever need to become active. (each datastore can grow up to 100GB in size).
I am also suggesting a qnap nas as an iscsi target to the VM's as a datastore which could be used for holding the archive mail if they want additional capacity to do that
The second site will have a typical Exchange install and hold just the passive copies of all of the datastores. The secondary MX will be here, this provides a disaster recovery solution should the main site be wiped out, but it is a manual process to make site B live - which I think is the right way of doing it....
Site A
2x Physical servers
Each with 1200GB of storage (total after RAID 10)
VM on each server with 50GB for OS, 50GB for queues (CAS/HUB role) these will then be load balanced with NLB
VM on each server with 50GB for OS, 100GB for logs, 800GB for EDB's (MBX role) this leaves an additional on each of 150GB if needed.
My plan is to split the mailbox stores into 4 on each mailbox server for load balancing, with enough space on each to hold the passive database copies if they ever need to become active. (each datastore can grow up to 100GB in size).
I am also suggesting a qnap nas as an iscsi target to the VM's as a datastore which could be used for holding the archive mail if they want additional capacity to do that
The second site will have a typical Exchange install and hold just the passive copies of all of the datastores. The secondary MX will be here, this provides a disaster recovery solution should the main site be wiped out, but it is a manual process to make site B live - which I think is the right way of doing it....
Why not just have them connect via Outlook Anywhere?