none
Two DAGs (Active/Active) in Two Data Centers With 50% Users Each

    Pertanyaan

  • Hi

    Our organization has decided to deploy fresh Exchange 2010 messaging environment so the planning process started some time ago. We have around 5000 users with following quota requirements:

    1. 2500 users @ 500Mb
    2. 1000 users @ 1024Mb
    3. 1000 users @ 2048MB
    4. 500 users @ 25000Mb

    Following design was finalized:

    1. Two Data Centers connected by redundant 2Gbps WAN links. Data Center 1 will be named as Site 1 and Data Center 2 will be named as Site 2 from the solution perspective.
    2. All the Exchange roles (HT, CAS and Mailbox) to be virtualized with all the 3 roles on single Virtual Machine.
    3. There will be two Database Availability Groups configured. DAG 1 will be configured with 3 multi role (HT, CAS & Mailbox) VMs, 2 placed in data center 1 and one placed in data center 2. Total of 2500 users (out of 5000) will be hosted in the mailbox servers that are part of DAG 1.
    4. Similarly, DAG 2 will be configured with 3 multi role (HT, CAS & Mailbox) VMs, 2 placed in data center 2 and one placed in data center 1. This DAG will host the rest 2500 users. So each DAG will host 50% of the total user population (so it will be an Active/Active user distribution model) and from the solutions’ perspective Site 1 will be primary site for users in DAG 1 and Site 2 will be primary site for users hosted in DAG 2.
    5. In each DAG, the 2 mailbox servers in primary data center will host 1 Active and 1 Passive database copy. The third mailbox server in secondary data center will host a lagged copy (no passive copies). The File Share Witness will be placed in Primary Site for each DAG.
    6. Hardware Load Balancers will be used to configure CAS Array in each primary site.
    7. Cisco UCS Blade B230 servers (Dual 2.13 GHz E7-2830 Processors, each with 8 cores so total 16 cores. 128 GB RAM) to be used as VMWare hosts.
    8. During the planning phase, the planning team decided to assign following configuration to each Exchange VM:
      1. 4 vCPU
      2. 32 GB RAM
      3. Sufficient Storage

    While all the above has already been decided and our teams’ job is to deploy the solution, we have some queries before we start it:

    1. The amount of RAM given is sufficient but we really doubt the virtual processor configuration decided for each VM. We’ve tried using the Exchange Storage calculator (V 18.2) to get an idea of what processor power is required but we can’t find the CPU Benchmark value for hardware model being used (Cisco UCS Blade B230 servers, Dual 2.13 GHz E7-2830 Processors with 8 cores) on the Spec.org website. Can someone please point us to some way of concluding whether the CPU allocation is sufficient for our case?

    2. What are the various options available for us from the Namespace standpoint? Can we have one single SAN Certificate for all the CAS servers and for securing all the services like OWA, Outlook Anywhere, ActiveSync and especially for Autodiscover?

    3. In the event when the Primary Site fails, and the DAG losses quorum (since FSW was also in Primary Site and now only 1 voter is available in Secondary Site), what steps do we have to take to ensure users in DAG 1 can access e-mails from Lagged copy placed in Secondary Site.

    Do we need to evict the lagged copy node from DAG or do we have to enable an alternate File Share Witness in secondary data center?

    4. Since we have a lagged copy (72 hours behind) in the secondary data center for each DAG, is the data loss sure in case the primary site goes down?

    5. Does Lagged copy mailbox server also contribute to Quorum voter?

    6. In the event when the Site 1 fails, what steps do we need to take to ensure CAS and HT access from the secondary data center?

    7. In this scenario, can we rely completely on Exchange Native data protection or is it suggested to have a VSS based Software/Hardware backup solution in-place?

    Any reference documents would be of great help.

    Regards


    zamn


    07 April 2012 12:22

Semua Balasan

  • 1. Depnding on what product you're using you can allocate 8 CPU.  However I would spec the VM like a physical server.

    2. You will need two names on your cert autodiscover.domain.com and mail.domain.com (common name).

    3. Datacenter Switchovers

    4. Not sure what you mean here, the lagged copied will have the transaction logs but wont be played until the threshold is met. 72 Hours, so you could replay these to bring up to date.

    5. Yes it does.

    6. They should be available, the only change you may need to make is to change the RPCClientAccess value to the corresponding site.  I assume each DC is an AD site therfore will have it's own CAS Array.

    7. http://blogs.technet.com/b/omers/archive/2010/11/05/designing-a-backup-less-exchange-2010-architecture.aspx


    Sukh

    08 April 2012 14:26
  • Hi Sukh

    Thanks for clarifying the points, just want to get some more insight on point no. 1. Since we can't find the CPU Benchmark on Spec.org website (which is required by the Storage Calculator sheet), how can we ensure that we allocate sufficient CPU cycle to mailbox server role?

    Also, how does Exchange 2010 (HT Role) ensures that the messages submitted to it by user are recovered in case the HT just fails after immediately accepting the message (i:e just before relaying it to next hop it fails for some reason and cannot be recovered).   

    Regards


    zamn

    09 April 2012 4:05
  • Hi Zamn

    1. I will need to look into the CPU question, but do you want to explain what benchmar you can't find?

    2. Exchange uses a feature called shadow redundancy, I'd suggest you digest this - http://technet.microsoft.com/en-us/library/dd351027.aspx


    Sukh

    09 April 2012 15:28
  • Hi Sukh

    As I shared earlier that we have the following hardware purchased for hosting Exchange 2010 VMs:

    Cisco UCS Blade B230 servers (Dual 2.13 GHz E7-2830 Processors, each with 8 cores so total 16 cores. 128 GB RAM). Each Exchange VM will be allocated following reqources:

    • 4 vCPU
    • 32 GB RAM
    • Sufficient Storage

    The amount of RAM given looks sufficient but 4 vCPU (virtual processor) configuration is really doubtful (We are colocating all the 3 Exchange 2010 roles on same VM). When we were trying to use the Exchange Storage calculator (V 18.2) to get an idea of what processor power is required, it asked for the following value:

    • SPECint2006 Rate Value

    The storage calculator suggests you to find this value at Spec.org, but we can’t find this CPU Benchmark value for hardware model being used (Cisco UCS Blade B230 servers, Dual 2.13 GHz E7-2830 Processors with 8 cores) in our case.

    Since this value is not there, we are looking for some suggestion on how we can ensure that our CU allocation is inline with our scenario.

    Also, the Shadow Redundancy feature helps HT in getting the feedback about the message in transit, so it can re-transmit the message if HT did not receive confirmation about the message delivery success. But my question is a little different, what I want to ask here is, how the Exchange 2010 environment ensures that a message submitted to HT is recovered, if HT server fails immedietly after receiving the message before it actually transmits it to next hop?

    Regards

     


    zamn

    10 April 2012 5:40
  • Hi

    Any inputs from people would be of great help.

    Regards

     


    zamn

    14 April 2012 9:40
  • In that case the message will be lost if shadow redundancy can't help. The likeless of that occurring is minimal if you ask me.  Redundant hardware etc...

    Sukh

    08 Mei 2012 10:06