Welcome to Exchange Team Blog Sign in | Join | Help

Syndication

This Blog

Updates to the Exchange 2010 Mailbox Server Role Requirements Calculator

==Version 4.5 Updates==

Enhancements

There are only three improvements to the calculator in this release:

  • Added the calculations to help you determine the minimum number of global catalog cores you need to deploy with your architecture.
  • Added the ability to define different RAID parity settings for your Restore LUN architecture.
  • Improved the calculation of the formatted capacity of a disk drive.

Bug Fixes

  • The following issues were identified and addressed in this release:
  • For HA solutions that chose to deploy in JBOD configuration, two fixes were made:
    • In certain scenarios, the calculator would calculate the number of mailboxes per database to be higher than the total amount of random IO available on the disk. As a result the calculator couldn't recommend a JBOD configuration. This has been corrected by rounding down the number of mailboxes per disk from an IO perspective calculation.
    • In certain situations the JBOD disk configuration isn't applicable. In past versions, the calculator would simply tell you this was because of "Insufficient Disk Capacity". This has been corrected to tell you if the insufficiency is based on disk capacity or IO capacity.
  • Addressed several CPU calculation related errors:
    • Fixed the issue where #VALUE was reported in the secondary datacenter processor core ratio table when you deployed only a lagged copy in the secondary datacenter.
    • Fixed the required mailbox CPU calculations to take into account certain site resilient scenarios where neither datacenter could support a single server failure (e.g. 2 member DAG).
  • Fixed an issue with the total disk count calculations with dedicated lagged copy servers as a result of rounding error.
  • Fixed an issue where /DAG LUN space totals didn't take into account the 2 LUNs / Backup Set architecture.

==Version 3.5 Updates==

Version 3.5 introduces the following fixes:

  • Improved the text on the input tab with regards to the number of database copy instances you would like for both HA and lagged copies.
  • Fixes an issue where in a high availability architecture the calculator may size the solution based on activating more database copies during a second server failure event than the total number of database copies deployed on the server.

==Version 3.4 Updates==

Version 3.4 corrects a memory and CPU utilization issue where you deploy a site resilient architecture with multiple mailbox servers and a single database copy in the primary datacenter.  Specifically, the calculator would determine the active database copy configuration after a single server failure and then size the CPU and memory requirements.  However, since there is only a single database copy in the primary datacenter, the solution cannot survive with all copies hosted in the primary datacenter.  Therefore, the copies need to be activated in the secondary datacenter.  Version 3.4 corrects this scenario by ensuring there are at least 2 database copies in the primary datacenter in order  to calculate the active database count after a single server failure.

==Version 3.2 Updates==

It's been a while since we discussed the Exchange 2010 Mailbox Server Role Requirements Calculator. Well I am pleased to say that today we are launching version 3.2 of the calculator.

This version includes the following improvements and new features:

  • Added processor core guidance for Hub Transport and Client Access server roles.
  • Added the ability to define a custom number of databases that you would like to implement in the solution.
  • Added support for 2-node site resilient Database Availability Groups.
  • Added 1 and 6 processor cores as selectable options.
  • Improved breakdown of the activation scenarios in a site resilient solution.
  • Improved breakout of the role requirements section.
  • The Storage Design tab now indicates that when you select a custom RAID configuration that the calculator ignores RAID-5 and RAID-6 for 5.xK and 7.2K spindles due to performance concerns.
  • Updated processor utilization results to show the processor utilization even if it is above the recommended threshold.
  • Made conditional formatting improvements throughout the calculator to warn you when you have a configuration that will not work.
  • Improved various cell comments.

This version also corrects the following bugs:

  • Fixed LUN Requirements tables to accurately reflect space requirements when database copies are deployed as each server may not host all database copies.
  • Fixed conditions that resulted in -1 lagged copies.
  • Improved the active database copies after first/second server failure calculations:
    • We now calculate and expose the worst case scenario (the server that has to host the most active databases) is used in sizing memory and CPU.
    • We now ensure that the secondary datacenter calculations only consider double server failures when there are 3+ HA copies located in the secondary datacenter.
  • Removed maximum memory stipulation in the minimum ESE cache memory calculation.

Hey where is Active/Active?

And for those that I know will ask, this version of the calculator does not include the Active/Active user distribution site resiliency scenario. For those that need that scenario, what I recommend is the following:

  1. Launch two versions of the calculator.
  2. Populate the first version for the first DAG in your design. This DAG (DAG1) will utilize Datacenter 1 as its primary location (and thus its user population is based out of Datacenter 1). It has site resiliency by having servers and database copies located in Datacenter 2 that can be activated in the event Datacenter 1 is lost.
  3. Populate the second version for the second DAG in your design. This DAG (DAG2) will utilize Datacenter 2 as its primary location (and thus its user population is based out of Datacenter 2). It has site resiliency by having servers and database copies located in Datacenter 1 that can be activated in the event Datacenter 2 is lost.
Datacenter 1 Datacenter 2
DAG1 Active Passive
DAG2 Passive Active

By implementing the architecture in this way, you can ensure that for the majority of scenarios except loss of datacenter, the users remain operational in their primary datacenter location.

Conclusion

Hopefully you will find this calculator invaluable in helping to determine your mailbox server role requirements for Exchange 2010 mailbox servers. If you have any questions or suggestions, please email strgcalc AT microsoft DOT com.

For the explanation of different tabs and how the calculator works, go here. Yup, we updated that too!

Finally, to get the new calculator - go here.

- Ross Smith IV

Published Friday, January 22, 2010 11:06 AM by Exchange

Comment Notification

If you would like to receive an email when updates are made to this post, please register here

Subscribe to this post's comments using RSS

Comments

 

Michel de Rooij said:

There's a small error in the description. In the active/active paragraph, the second bullet reads:

* Populate the first version for the first DAG in your design. This DAG (DAG1) will utilize Datacenter 1 as its primary location (and thus its user population is based out of Datacenter 1). It has site resiliency by having servers and database copies located in Datacenter 2 that can be activated in the event Datacenter 2 is lost.

This should be "in the event Datacenter 1 is lost".

Keep up the good work!

Regards,

Michel de Rooij
January 22, 2010 8:33 PM
 

Exchange said:

Thanks for catching that, Michel.

Ross
January 23, 2010 9:23 AM
 

bday said:

Nice job on the update, Ross! :)
January 23, 2010 11:19 AM
 

Robert said:

OT, but are you going to have a post on Exchange Server 2007 SP2 Update Rollup 2?  I'm having an issue.
January 25, 2010 4:10 PM
 

Exchange said:

Robert, for any issues you experience with E2007 SP2 RU2, please either open a support case or start a thread on our Support Forum - http://social.technet.microsoft.com/Forums/en-US/exchangesoftwareupdate/threads/

Ross
January 25, 2010 4:42 PM
 

jdbizzy said:

great catch!
January 27, 2010 4:31 PM
 

scottie37trevino said:

"OT, but are you going to have a post on Exchange Server 2007 SP2 Update Rollup 2?  I'm having an issue."

I too am struggling with this
January 28, 2010 10:11 PM
 

suman.san14 said:

Hi,
I got an error while running Jetstress 3.0 for Exchange 2010 testing, on Windows Server 2008, of "WaitHandles must be less than or equal to 64"....

can anybody help how to fix that.....

regards,
suman
January 28, 2010 11:25 PM
 

Dillon said:

In a site resiliency deployment (active/passive, two servers in each site, total HA copy instances of 4) if the inputs for number of servers hosting active mailboxes/dag = 2 and  number of HA database copy instances deployed in secondary datacenter = 2, the Role Requirements page returns a number of divide by zero errors. It does not seem to affect other dual-site configurations with other numbers filled in for these two values.
February 5, 2010 1:41 PM
 

BR said:

I have a question plz ,
the Database Space Required / Replica calculation result is per what ? Day , month , yeah ?!
February 17, 2010 6:10 AM
 

Exchange said:

BR - there really isn't a timeframe.  I guess worst case, it is the amount of space required for the lifetime of the storage assuming the inputs you chose remain the same.

Ross
February 20, 2010 4:34 PM
 

iamme said:

Odd, when I specify that I have 3 servers in my single site, it tells me to set MaximumActiveDatabases to let's say 30 and that I could support 1 server failure.  But if I then specify I have a second site, suddenly my main site allows me to support up to 2 failures in just my primary site and then my MaximumActiveDatabases shoots up to support all of my databases (let's say 3).  Not sure why that happened and seems like a bug.

Maybe in the future we can say how many server failures we'd like to account for in each datacenter.
February 20, 2010 10:12 PM
 

Exchange said:

Folks, I wanted to let you know of an issue in the v4.5 calc - if you are designing configurations to support multiple DAGs, there is an issue in the reporting of the number of processor cores you need for the various roles.  basically the calculator is making an assumption of only having a single DAG in the environment.  So you will need to multiply the processor cores per role by the number of DAGs you are deploying to see the full number of cores required.  This will be fixed in the next release.

Ross
February 27, 2010 1:19 PM
 

Exchange said:

iamme -

I'm actually improving the reporting of this in the next release.  Essentially when you move to a multi-site architecture, you support the ability to have cross-site *over events.  So you could survive two server failures (in certain situations within the primary datacenter) because some of the databases can be activated in the secondary site.

Ross
February 27, 2010 1:21 PM

What do you think?

(required) 
(optional)
 
Enter the code you see below

(required) 

News


This blog and its contents are provided "AS IS" with no warranties, and they confer no rights. Use of any included script samples are subject to the terms specified in the Terms of Use.
New! Would you like to suggest a topic for the Exchange team to blog about? Send suggestions to us.

Exchange Server 2010 - Get the Release Candidate



Poll:

Other Exchange Blogs from MSFT