Load Balancing Exchange Server 2010 and 2013

Summary: In this session we will have a look at how we can load balance our Exchange 2010 and Exchange 2013 environment. First we look at the some load balancing essentials where I will discuss which terms are important. Once we completed this part we will have a look at how to configure your load balancer for Exchange 2010: which configuration options are available,  which things must  be configured on the Exchange side and followed by a demo where I will let you see how to do it. As final part of the presentation we will look at Exchange 2013: what are the differences, which things are important to keep in mind and a demo where I will let you see how to do the load balancer configuration.

About Johan Veldhuis

Johan Veldhuis is an Exchange Server MVP and Technical Consultant who works at a consultancy firm which is specialized in Microsoft Unified Communications solutions. In that role he is responsible for designing, implementing, migrating and troubleshooting Microsoft UC solutions. Besides his own blog on his website, Johan has been a regular author for Simple-Talk and several other blogs, and is also a member of The UC Architects which is a bi-weekly podcast covering both Exchange and Lync related topics.

Comments

  1. Patrick Boshell says:

    amazing video thanks for the time and effort you put into this

  2. Nicolas Picard says:

    Very nice video. Thank you for these valuables information

  3. Raúl Silis says:

    Is it possible to use Windows Network Load Balancing in Windows 2012 with Exchange 2013?

    • Johan Veldhuis says:

      Hi Raul,

      Yes it is supported but there are somethings you must keep in mind:

      - You can’t use WNLB on Exchange servers which are a part of a DAG. If you want to use WNLB you need to have separate servers for the Client Access Server role and the Mailbox server role;
      - WNLB only checks if an IP is reachable so if a service crashes on the server but it is reachable via IP it will still get traffic offered which results in lose of functionality. In this case you will need to manually remove the server from the WNLB
      - Using WNLB can result in port flooding which has a negative effect on your network
      - WNLB uses source IP address as affinity this result in a less effective solution when the source IP pool is small (think of remote networks or when using NAT)

      Hope this answers your question.

      Johan

  4. Chris Phillips says:

    Hi Johan,

    I’m concerned that the description & video/flow-diagram of DSR traffic is inaccurate or worse, that my understanding of it is hugely flawed!

    I was under the impression that with DSR, inbound traffic hits the Load Balancer, then gets passed to the real server(s), but that the Return traffic goes Directly to the client, not via the Load Balancer. This is why the real servers require the non-ARPing loopback VIP address configured.

    Please can you clarify this?

    Regards,
    -
    ChrisP

    • Johan Veldhuis says:

      Chris,

      Your completely right about DSR but the diagram does display a workflow where DSR is not applied. In that case all traffic will be send back via the load balancer.
      If you decide to use DSR the traffic from the server is send directly back to the client as you already mentioned.

      Hope this clarifies your question.

      Regards,
      Johan

      • Chris Phillips says:

        Thanks Johan,

        Phew, my mind is at rest again :)

        Poor old DSR gets a bad reputation. On the web, I see lots of focus on the cons and not so much on the pros.

        We use DSR on our Linux-based mail servers, which were in a very well established network. Because we could not afford to interrupt the service in order to have them load-balanced in a 2-armed arrangement, DSR was perfect for us and works tremendously well.

        We introduced load-balancers on the same network segment & configured them for DSR, added the non-ARPing loopback VIPs onto the live mail servers and switching to a load-balanced service was as simple as a DNS change. While the DNS propogates the old service continues to work and also if there was a catastrophic failure of the load-balancers, we could switch back to a non-load-balanced service with a simple DNS change.

        We also use the same load-balancers in a 2-armed arrangement, to balance Microsoft Exchange 2010 & 2013. The best of both worlds.

        Thanks for your video. BTW, Kemp also now have templates for Microsoft Exchange 2013 :)

        Cheers,
        -
        ChrisP