As some of you might know, I work as a Technical Marketing Engineer for the Converged Infrastructure Group at NetApp. As part of my day job, I work on designing reference architectures for FlexPod. FlexPod is a predesigned, best practice data center architecture that is built using the Cisco UCS servers, Cisco Nexus family of switches, and NetApp All-Flash FAS (AFF) A-Series systems. The recent architecture that I was working on was focussed on solving a specific use case, “How can I run Microsoft Exchange 2016 and Microsoft SharePoint 2016 in my Datacenter”.
I am aware of the fact that many of you are thinking about moving to Office365 if you haven’t moved already. If that’s the case then this particular reference architecture is not for you. This architecture caters to a very specific group of people who want to continue owning and running their Email and SharePoint services on-prem due to security, data locality or other reasons. If you identify with this group of people then read on!
Glenn Sizemore and I had been working on this reference architecture over the last quarter. This design helps you build out an infrastructure using FlexPod that is able to support 10000 active Exchange and 10000 active SharePoint users simultaneously. We used the latest generation of NetApp Controllers running ONTAP 9.1 along with Cisco B200 M4 Series blades and Cisco Nexus 9396PX switches. Running Exchange and SharePoint 2016 on ONTAP 9.1 has the following benefits:
- Increased data availability
- Simplified and Centralized data management
- Efficient storage utilization
- Highly scalable
- Reduced Costs
- Enable Multi-Tenancy and QoS
The following diagram shows the logical topology diagram for the design:
We had a FlexPod running at the primary site which was using the latest NetApp hardware, and a FlexPod running at the secondary site which was using older storage controllers. For running the Exchange 2016 workload, we were using eight Mailbox servers with 4 databases each. Each Mailbox server was responsible for supporting 1252 users. NetApp AFF A300 on the primary site was hosting the LUNs for primary and secondary mailbox copies. NetApp AFF 8040 at the secondary site was hosting the LUNs for the tertiary mailbox copies.
For SharePoint 2016, we were running multiple Web Front End, Application and Search servers to be able to serve 10000 users and also tolerate any node failures. SharePoint workload was using SQL 2016 Databases on the backend. To make the databases highly available, we were using Database Availability Groups(DAG) with a synchronous database copy on the primary site and an asynchronous database copy on the secondary site. If you want to know more about the architecture and the design decisions that went into the design, please refer to the following two NetApp Verified Architecture documents:
As part of this design, we have also included applications for Data Protection. We are using the NetApp SnapManager for Exchange and NetApp Single Mailbox Recovery Applications to protect Exchange data. And we are using DocAve Manager v6 to protect the SharePoint data. Having these applications enables us to provide backup and recovery capabilities as part of the design, which is quite important.
In addition to the design pieces of this NetApp Verified Architecture, we also went through a series of performance and load testing to make sure that the design that we are proposing is able to run without any hiccups when put under load. If you refer to the two guides that I have linked above, you will be able to look at the test criteria and the performance graphs from those test runs. To conclude, I can say that FlexPod Datacenter with NetApp AFF performed well, reaching a combined peak IOPS of over 10K while averaging 15% CPU utilization during most operations. All test categories demonstrated that, based on the 10,000-user load, the FlexPod Datacenter with AFF A300 system is capable of supporting up to 100,000 users while still being able to perform appropriately even in the event of a storage, host, or network failure.