Major update to the Mailbox Server Role Storage Requirements Calculator – v 11.0

Please go on our Mailbox Server Role Storage Requirements Calculator updates tracking page to see what is in this new version!

A blog post explaining the calculator (updated for this new version) is here.

Comments welcome!

- Ross Smith IV

Comments (19)
  1. Dillon says:

    Excel sheet still says version 9.6 and has no updates on the update history tab. Yes, the filename says v11, but the calculator appears no different once it has been opened in Excel.

    Messed up the upload perhaps?

  2. Thanks for reporting the issue, Dillon.

    We’ll get it fixed ASAP.


  3. Exchange says:

    Dillon – thanks for reporting this, I have just corrected this. I messed up when uploading the file. Sorry about that everyone!

  4. Dillon says:

    Thanks for the update. :)

  5. bday says:

    Hello. I’ve not yet had time to try out the calculator, but when I do will it aid me in estimating bandwidth requirements between two geoclustered locations when using CCR? I may not word the following correctly, so please bear with me. :)

    I do not yet know how a mailbox role server chooses a hub transport role server. Is it round robin? Chooses one then sticks with it? Do the mailbox role servers talk to each other and split u pthe HTs dynamically?

    We will have HT role servers at both datacenters and one node of an active/passive CCR cluster at each datacenter. These datacenters will probably be part of the same dedicated Exchange active directory site as we have centralized Exchange in our org in its own domain. Is it possible a mailbox role server could choose a HT role server in the remote datacenter to send mail through even for users on the same message store? If this is possible, how do I calculate the necessary nearly-sustained traffic between the physical sites?

    Thank you. :)

  6. bday –

    Yes the calculator does provide an estimate on log replication throughput necessary to sustain replication.  It does not provide guidance on other factors (like client connectivity, AD replication, transport, additional apps, etc).  This number is derived from the log traffic that is generated per hour in the day, which is something you will need ot measure in your existing environment.  If you cannot obtain that data, then you can estimate it, but beforewarned, that it’s just an estimate and thus thelog replication throughput metric may be significantly off when you implement in production.

    Yes mailbox servers will round robin requests between multiple hub transport servers within the same AD site.  No mailbox servers do not communicate to other mailbox servers which hub servers they are using.

    For more information on geographically dispersed planning, I would take a look at –


  7. bday says:

    Thank you, Ross. So without splitting Exchange into 2 AD sites (one per datacenter) it is possible I could generate excessive and unecessary WAN traffic by a mailbox server in datacenter 1 choosing an HT server in datacenter 2.

    I’ve been trying to get baseline performance analyzer working in our 2003 environment, but it seems to just timeout. Is this what I should be using to get my current log traffic numbers? At worst I could snapshot the amount of logs in a 24 hour window (after data backup to just before data backup) and also the amount during business hours and make some estimations.

    Thanks again.

  8. bday says:

    Took a spin through the calculator, and saw the Collectlogs.vbs file, that will come in handy!

    Is there any place to input how many exchange servers you in the exchange org or should I take some of the output (bandwidth requirements in this case) and multiply them by the # of servers in the org? Is the grown pretty linear as servers are added assuming each server has a similar profile of users?

    We’ve got 4k users per server over 8 (soon 10) servers so I’m trying to find what the collective infrastructure needs are. Thanks! :)

  9. bday,

    Yes the calculator calculations are for a single server only, so you will need to take the output and multiply that by the number of servers.

    As for the question regarding using HT servers located in other sites.  A Mailbox server will not communicate with HT servers outside of its AD site.  So in a geographical dispersion case, you have two options, span the AD site, or use separate AD sites (also note with Windows 2003 Failover Clustering you have to span the subnet).  As to which model you use depends on a few factors and each has some issues which are discussed in the link I listed earlier.  If you do go the span AD site sceanrio, then all MBX – HUB communication will occur over the WAN (via RPC) so you have to make sure that latency etc does not affect that communication because by using RPC we assume a well connected network (aka LAN).  As to how much traffic that would generate and the resultant throughput you would need, I couldn’t say.

    The other tool you may want to look at is Exchange Profile Analyzer.  It will help you collect the message profile, message size and other things the calculator asks.

  10. bday says:

    Thanks again, Ross, much appreciated. Yes, we are presently planning to span the Exchange VLAN, subnet, and AD site to the 2nd datacenter over a 1Gbps (increaseable) fiber link. The latency I’ve been told by our network engineers should be next to nothing, we’ll see how much water it holds one we’re in productoin. :)

    The MBX – HUB communication over the WAN link will just have to be a "wait and see" for now I guess.

    I have the Exchange Server Profile Analyzer running right now. Its a long weekend so I’ve got about 3.5 days to let it churn away on our servers. I’m trying to see if it can finish a week’s worth of data within that timeframe. I had trouble getting the GUI version to run, but the commandline version seems to be doing its thing pretty well!

    Thanks again.

  11. Stefan says:

    hi, the contribute has helped me, my friend fales to correct. I hope you will write furthermore good article.

  12. bday says:

    Quick question.  Can collectlogs.vbs use UNC paths, or must they be local paths?

  13. Yes it uses local paths, though you could modify the source code to allow for UNC paths.


  14. bday says:

    Ross, thank you very much. Can I bug you once more?

    I’ve collected the # of transaction logs created per day and per hour for all of our servers. Should I be make 10 separate calculator files (one per exchange server), input the logs created for only that one server, enter 4,000 Tier-1 users, and then add the final bandwidth values together from each calculator, or should I be creating one calculator file with the cumulative sum of all log files created across all servers?

    If I enter the log files created on one server with 4,000 Tier-1 users (what each of our severs hold), the required bandwith reads as 16.46Mbps. If I leave the Tier-1 user value alone, go back and enter the cumulative values of log files created on all of our servers, the number goes up to 17.10Mbps.

    If I change Tier-1 users from 4,000 to 40,000 (our projected cumulative # of users across all servers), the required bandwidth gets multiple by a factor of 10; 164.60Mbps and 171.00Mbps respectively.

    I don’t understand this, can you help me out? Transaction files are 5MB (Exch 2003). Why does the # of users affect the required bandwidth if the total # of transaction logs created are the same at the end of the day?

    Thanks in advance!

  15. bday,

    The numbers you enter for the transaction log generation table are used to determine the log generation percentage that occurs per hour.  This percentage is then multiplied with the calculated log files that are generated per day (Average Transaction Logs Generated / Server / Day) to determine the throughput required based on what we believe you will be generating in terms of transaction logs.  The metric "Average Transaction Logs Generated / Server / Day" is based on user generated logs per day (which depends on message profile, message size, and number of users) and mailbox moves per day (assumes that the weekly mailbox move percentage is actually performed across the entire week).


  16. bday says:

    Thank you for your patience, Ross. :) What do you do if EPA data shows your users averaging 11 received and 13 sent (7 send, 4 reply, 2 forwarded) messages per day? All of the drop down options are weighted more heavily to receiving messages than sending.

  17. Bday,

    I would go with a light or average profile.  If you measure over a period of time (week or two and across a few servers) you may find the profile bend more to one or the other.


  18. TBerger says:

    Is the # of logs generated per hour based on 5 MB log files or 1 MB log files?  we are consolidating 4 2003 servers into 1 2007 server and am having a hard time predicting bandwidth usages as a result.  Should I sum the log files per server per hour [currently 5 MB log files] to get this value?



  19. Hi Tim,

    Yes you can use your existing server environment to calculate the number of logs generated per hour of the day.  What we do with this data is then determine the percentage of logs that are generated per hour (basically we are assuming that your activity isn’t changing when moving to Exchange 2007).  This percentage is then multiplied by the total logs the Exchange 2007 server will generate based on the other input factors you selected.


Comments are closed.

Skip to main content