none
How to approach sizing new Hub Transport load?

    Question

  • Hi All

    I've asked around this question a bit recently, so forgive me if this covers similar ground.

    I have 2 HT servers (CAS and HT dual role), inherited when I started 4 weeks ago. The email receive load internally seems fairly consistent in July\August at about 65000 emails per day across these servers. This is for 4000 active mailboxes, so actually quite a low usage profile I'd say, looking at "about" 30 000 emails per server per day (The other 4000-5000 explained below).

    However, of that 65000, about 4000 a day is from one Application server, at random times (Aside from a consistent morning burst), using just 1 sending Email address. These 4000 emails are NOT load balancing, they are connecting to 1 HT server largely, and then completing that session by sending all the 1000's through it. The "virtual SMTP relay" address for the Hubs is a virtual address, both Hub servers are associated to it via DNS alias.

    The above load is going to increase by October 2017 to around 6000 a day for this one email address (And on some days about 12000 a day for 1 week), and then in March 2018 to a consistent 12 000 a day for this email address.

    I am telling the Department that sends these emails that they need to outsource by March 2018 to a third party, otherwise I'll have to start considering dedicated HT servers and Receive connectors as well as queue delays for quiet hours, and this is a responsibility my IT department does not want to accept for this internal department.

    HOWEVER, in the meantime, what options do I have to determine if the 6000 emails per day (plus the 12 000 per day bursts for a 5 day period) can be handled by my existing setup come October 2017?

    It feels like on the 12 000 emails a day bursts, just one HT server would be taking all that load because there is no Load Balance solution in place for emails from an internal App server, and that App server cannot send to multiple hosts unfortunately.

    I have run the std Exchange and system perf counters over a 2 day period last week, between 5am and 8am (when an existing 3000 email run happens), and there doesn't appear to be any issues with server performance, everything is under the 50% threshold for the respective counter (RAM, CPU, RPC connectivity, SMTP send queue and so on).

    What I'm trying to determine is how I can best judge whether a single HT server can handle, lets say 42 000 emails per day, (up from 32 000), with those 10 000 emails potentially being sent as 1 continuous run...... The internet pipe we have is vast, so I'm not convinced that will be a bottleneck, but I don't want an entire HT server to be creating send delays for other departments by virtue of that Hub server processing 10 000 emails of around 200KB in size each. The Hub server is a VM, 4 VcPu and 12GB RAM.

    Can I have some guidance, maybe based on some real-world experience you guys have, of a HT being able to handle 45 000 or so emails per day, but with a single email run that might be of the 10 000 email size....?

    Options I'm considering are:

    1.) We'll be ok in October, 45 000 receive emails processed a day for an HT is not vast, even with a 10 000 email burst)

    2.) Create a scoped receive connector for this sending app server, delaying the sending of emails until the late evening (I have concerns though about emails being queued on disk during the day, not because of size, but the risk of losing them if something breaks)

    3.) a standalone HT server built for them, with a Receive connector to only accept emails from that server (And *somehow* preventing the other MBX servers from using this HT. I know this can be done, I'd just have to figure out how).

    The question I'm trying to answer is, how can I determine, through some checking, if the simplest approach (Do nothing) will be good judgement?

    Coop

    Monday, August 28, 2017 2:11 PM

Answers

  • Honestly, if you have sized your transport hubs correctly, it should be fine. Exchange will throttle if it needs to. 

    Looking at those specs, it looks fine. I have something similar but 8GB and can handle more than that with bursts...

    • Marked as answer by Coopie11 Tuesday, August 29, 2017 2:22 PM
    Monday, August 28, 2017 10:26 PM

All replies

  • Another Option I have, is something Allen Wang suggested some weeks ago, and to use a separate IP address and Network interface for email FROM the source APP server, however, I don't know if that will alleviate problems the exchange sub-system has in processing emails during a 10 000 email run, I suspect internal HUB queuing would still be an issue.......I don't know.

    Same guidance would still be appreciated, how best to determine if the Load spikes I mentioned in post 1 can be tolerated, or whether I do need to add in a mechanism to deal with them for October 2017.

    Monday, August 28, 2017 3:02 PM
  • I appreciate that my first post was very long, but anyone had the patience to read it and have some insight :-)?

    Coop

    Monday, August 28, 2017 9:48 PM
  • Honestly, if you have sized your transport hubs correctly, it should be fine. Exchange will throttle if it needs to. 

    Looking at those specs, it looks fine. I have something similar but 8GB and can handle more than that with bursts...

    • Marked as answer by Coopie11 Tuesday, August 29, 2017 2:22 PM
    Monday, August 28, 2017 10:26 PM
  • Hi,

    It should be ok. It’s recommended to refer to the following blog for sizing and capacity planning guidance:

    https://blogs.technet.microsoft.com/exchange/2013/05/06/ask-the-perf-guy-sizing-exchange-2013-deployments/

    Hope it helps.


    Regards,

    Jason Chao


    Please remember to mark the replies as answers if they help.
    If you have feedback for TechNet Subscriber Support, contact tnmff@microsoft.com.

    Tuesday, August 29, 2017 4:36 AM
    Moderator