Microsoft Exchange on Nutanix Best Practice Guide

I am pleased to announce that the Best Practice guide for Microsoft Exchange on Nutanix is released and can be found here.

For me deploying MS Exchange on Nutanix with vSphere combines best of breed application level resiliency (in the form of Exchange Database Availability Groups), infrastructure and hypervisor technologies to provide an infrastructure with not only high performance, but with industry leading scalability, no silos and very high efficiency & resiliency.

All of this leads to overall lower CAPEX/OPEX for customers.

In summary by Virtualizing MS Exchange on Nutanix, customers realize several key benefits including:

  • Ability to use a standard platform for all workloads in the datacenter, thus allowing the removal of legacy silos resulting in lower overall cost, and increased operational efficiencies.
    • An example of this is no disruption to MS Exchange users when performing Nutanix / Hypervisor or HW maintenance
  • A highly resilient , scalable and flexible MS Exchange deployment.
  • Reducing the number of Exchange Mailbox servers required to maintain 4 copies of Exchange data thanks to the combination of NDFS + DAG. (2 copies at NDFS layer / 2 copies at DAG layer)
  • Eliminate the need for large / costly refresh cycles of HW as individual nodes can be added and removed non disruptively.
  • Simplified architecture, no need for complex sizing architecture or risk of over sizing day 1, start small and scale VMs, Compute or storage if/when required.
  • No dependency of specific HW, Exchange VMs can be migrated to/from any Nutanix node and even to non Nutanix nodes.
  • Full support from Nutanix including at the Exchange, Hypervisor and Storage layers with support from Microsoft via Premier Support contracts or via TSANet.
  • Lower CAPEX/OPEX as Exchange can be combined with new or existing Nutanix/Virtualization deployment.
  • Reduced datacenter costs including Power, Cooling , Space (RU)

I hope you enjoy the Best Practice guide and look forward to hearing about your MS Exchange on Nutanix questions & experiences.

10 thoughts on “Microsoft Exchange on Nutanix Best Practice Guide

  1. Pingback: » Exchange Best Practices and 1.4Million Mailboxes on Nutanix NX-8150 Long White Virtual Clouds

    • MS official policy currently states NFS is not supported, however we are hoping MS will change the stance on this in the future. In the mean time, Nutanix can provided supported solutions via iSCSI or SMB 3.0 in addition to Nutanix providing End to End support for MS Exchange, the Hypervisor and the Nutanix platform.

      Nutanix is also a member of TSANet and can work with Microsoft directly on any support issue.

      I would recommend any production deployments of Exchange purchase Premier support which will also ensure MS provide support regardless of underlying storage protocol.

  2. Hi Josh,

    Nicely written best practice.
    I have a question/remark concerning the container settings.
    I’m referring to the example on the top of page 32, which would be exactly my environment, and page 23 where only delayed compression and no deduplication is supported. If i’m correct, a 2TB mailboxes environment would result in a bruto 8TB consummation of your storage pool. To start your passive DB’s are exact copies of your active copies, so in post dedup terms this should give you a bruto 4TB profit. Next to that we know that since Exchange 2007 we no longer have single file storage inside the EDB’s. So if Nutanix would be able to recognize the equal ‘parts’ inside the EDB, you would be dedupping that DB as well. (no idea if this would work).
    You state that the DB’s change often, but inside 90% of mails are archive…non changing bits.
    I have my own Nutanix cluster and despite the good perfomance and managment, pricing is not low budget… so using the storage wisely is recommended.
    Could you share your thoughts/experience on using post deduplication.
    (specially keeping in mind that 90% of the DB doesn’t change….old mails)

    thx,
    Bjorn

    • Hi Bjorn,

      In line with Microsoft’s recommendation to keep 4 DAG copies for Exchange, Nutanix recommends 2 DAG copies + Resiliency Factor 2, which equates to 4 copies.

      While we can deduplicate this data, again in keeping with Microsoft’s recommendations, and the purpose of a DAG (to provide higher data availability) we recommend not to use deduplication, and to instead use Compression. This means we always keep 4 copies of the data, but you will have capacity savings from the compression.

      I agree the vast majority of an Exchange mailbox is generally cold data, and as such, does not change, nor need high performance.

      Nutanix ensures this cold data is stored in the SATA tier to reduce the cost per GB of the storage.

      We are currently developing enhanced compression and data protection techniques which will result in further capacity savings, so stay tuned.

      Thanks for your comment.

  3. Hi Josh,

    Thanks for the feedback, if you don’t mind…I wish to discuss this a bit further.

    As a starter, can you somewhere show me that recommendation of Microsoft to keep 4 DAG copies? The only one I know is when you want to go ‘backupless’ in an environment, you need to have +3DB copies. (along with circular logging, etc..).
    But I don’t see the point of replicating a passive DB on storage level.
    Today we have 2 SAN’s, both host one active DB and one passive DB.
    This gives me maximum HA/Res + minimum storage waist. Bringing this to my nutanix cluster (3 blocks in 3 seperate buildings with 10GB links) I would maintain the same design. Due to RES factor 2 it would automatically double the necessary space. At that point I would guess that post Dedup would come in handy to safe the space. i see no disadvantage towards Microsoft in this.
    The main question is, will Nutanix be capable of dedupping it!

    I don’t disagree on written best practices here, but I often find myself thinking ‘Are they giving away storage for free?” Know that we’re an European based company and Europe has a whole other mindset on sync Data / copies compared to US mindset(east coast/west coast DC’s).

    If you wish to discuss this outside your blog, no problem for me. It’s always interesting to hear other peoples opinions!.

    Regards,
    Bjorn

    • Hi Bjorn,

      Your correct, MS recommend greater than 3 copies when going backup less (a.k.a Exchange Native backups).

      When using backups, 3 is generally what I see recommended around the industry by Exchange MVPs but it will vary based on customer requirements and constraints.

      On Nutanix, I only recommend 2 DAG copies with an optional LAGGED copy as we already provide data protection at the NDFS layer.

      Currently Nutanix recommends Compression, not deduplication as deduping data which is deliberately duplicated at the application layer may not be advisable.

      With compression, depending on the email profile, you can get significant savings.

      One of the good things about Nutanix is the new email is written too and read from SSD, so the user experience is great, while the cold mailbox data (especially for larger mailboxes) is tiering off automatically to SATA, so the cost / GB is much lower.

      As a result, the fact we use RF2 to provide data protection, is fine even when using a DAG with 2 copies.

      But yes we could dedupe the data.

      Happy to discuss in more detail if you wish.

      Good questions BTW!

      Cheers

      • Hi Josh,

        Thx for the reply.
        Depending on your requirements, for me DAG with 2 copies combined with Nutanix’s RF2 provides enough data protection. A lagged copy could be considered in case you wish to recover from a ‘virus/malware’ while working backupless. I also agree on the compression/dedupe thoughts. While technically dedupe would work, you still need to take in count that the dedupe is done by a 3rd party (Nutanix) from Microsoft’s POV. So if something goes wrong with the DB’s due to the dedup (afterall, everyone can make a mistake), the impact could be big. Optimal would be having the lagged copy in a seperate container, combining best of all worlds.

        Was nice having this conversation.
        Keep up the good work.

        Bjorn

  4. Pingback: Nutanix Platform Link-O-Rama | vcdx133.com

Leave a Reply to Josh OdgersCancel reply