VDI Reference Architecture Monthly – November 2016

Here we are again with another post for the Reference Architecture Monthly. I’m finding that there is a lot of great information with respect to proofs of claim and sizing, which is incredibly useful if you are trying to architect your own solution. In most cases these documents will carry this information, but what is more interesting is that there is almost always a unique morsel of information, sometimes hidden in the context of some great testing and analyzation. So, I’m going to start focusing on what I consider “the cool bits”.

VMware Horizon 7.0 and VMware vSphere 6.0 with EMC Unity Storage Systems

by EMC End-user Computing


This paper was a reference architecture summary for EMC’s End-User Computing VMware Horizon 7.0 and VMware vSphere 6.0 with EMC Unity Storage Systems Solution Guide. I’m sure the Solution Guide has a lot of the details I typically look for in these reviews. With that said, it is clear that the goal of this document is to highlight the benefits of EMC’s new Unity Storage Systems—which were released earlier this year—and their capabilities with respect to VDI performance.

In this paper, they review 5 reference configurations using Windows 7, Windows 8.1 and Windows 10, leveraged both Login VSI Office and Knowledge workers and used both VMware linked-clones and instant-clones.


  • Users:  Up to 2,400 Office and Knowledge workers
  • Provisioning system: VMware Horizon 7.0
  • Desktop OS: Windows7, Windows 8.1 and Windows 10 (Office version unknown)
  • Hypervisor: VMware vSphere 6.0
  • Hardware: Cisco UCS M3-200 and M4-200 were used for compute while storage was represented by the new Unity 300 and Unity 400 systems, while higher user counts were extrapolated based upon test data using the Unity 500 and Unity 600.

The cool bits

One of the things I found particularly interesting in this document were the I/O performance profiles between linked-clones and the relatively new instant-clone technology. There are definitely huge benefits to using instant-clones if you want to spin up a pool of desktops extremely fast, but I’ve always been curious about the performance effects on the system when using instant-clones, especially when considering storage. Check out the following table where there is 2000 IOPS difference for the 700 users in test.

vdi reference architecture monthly november 2016 virtual desktop density

Reference Architecture for Workloads using the Lenovo ThinkAgile CX Series Platform

by Lenovo


Lenovo offers a complete package in this reference architecture with their converged infrastructure called the ThinkAgile CX Platform. This one is interesting because it is a collaboration between Lenovo and Nimble Storage and has everything you need to get started with a VMware Horizon VDI solution—besides AD, DHCP/DNS and Licensing Servers–and what’s better than a solution that has already been tested for you.


  • Users: 1448 stateless linked clones and 1502 persistent full clones
  • Provisioning method: Both persistent and stateless methods, full and linked-clone respectively
  • Provisioning system: VMware Horizon 7
  • Desktop OS: Provide guidance on Windows 7
  • Hypervisor: VMware vSphere 6.0 Update 2
  • Hardware: CX4200 including Lenovo x3550 M5 servers, Lenovo networking and Nimble CS3000

The cool bits

The information I really enjoyed comparing was the performance differences between stateless, linked-clones and persistent full-clones. To start, Lenovo looked at the differences in densities per server for Office Workers and Knowledge workers, each with different results depending on whether they were stateless or not. See here:

vdi reference architecture monthly november 2016 virtual desktop performance

Also, look at the difference is storage transactions between the two. I’m a big advocate for focusing on write IOPS and latency in a storage system. This is because writes are the most sensitive operation in that it carries an overhead depending on the redundancy type, it must be cache protected, compressed and typically is the most stressful operation with any storage subsystem.

vdi reference architecture monthly november 2016 performance with different login rates

There are lots of great details around storage performance in this paper, so I suggest you check it out. There is also some sizing guidance which takes into account the available capacity each node should have in case there is a failover and the desktops from the failed node need a new home amongst the other nodes in the vSphere cluster.

VMware Horizon 7 on vSphere 6.0 Density and Performance Test Results Baseline for Windows 10, Office 2016 and Login VSI 4.1.4 test results

by DELL Cloud Client-Computing Engineering


In this paper DELL looks at a number of differences between systems including Login VSI versions, Windows desktop OS, Microsoft Office, and screen resolutions of the virtual desktops on a single server. There is a bunch of interesting data which can help in determining server density based on these variables.


  • Users: 270 Task, 180 Knowledge and 150 Power Worker per server
  • Provisioning method: Both persistent and stateless methods, full and linked-clone respectively
  • Provisioning system: VMware Horizon with View 7
  • Desktop OS: Provide guidance on Windows 7 and 10
  • Hypervisor: VMware vSphere 6.0 Update 2
  • Hardware: DELL PowerEdge R730, Intel Xeon CPU E5-2698 v4 @ 2.20GHz, 512GB RAM, and locally attached SSDs.

The cool bits

If you are looking for a comprehensive guide as to how many desktops per server you can get, depending on the types of workers you have, then this guide will be helpful to you. While VSImax was not reached for any of the user types and capacities tested, they do get pretty close to using up the system resources from the vSphere host’s point of view, but they also note something very important with respect to the BIOS. When it comes to setting up the server BIOS it is critical that the performance profile is selected as it enables much more headroom in the processor performance. Believe it or not, most of the time in our support cases we can tell if the BIOS is not set to the performance profile because of the way the VSImax graph looks.

One other notable bit of data that I haven’t seen very often in these reference architectures is the impact on the system resources depending on the resolution chosen. To please modern users on current monitors, you need to support a pretty high resolution. In this case DELL tested 1280×720 resolution and 1920×1080.

In DELL’s words:

“The graphs below show the significant increase in CPU and memory requirements associated with the increased display resolution. CPU utilization increased as much as 14% while memory usage increased by 27GB. IOPS and Network utilization were unchanged.”

vdi reference architecture monthly november 2016 cpu usage

QNAP Enterprise Storage All-flash Solution for 500 VDI seats with VMware Horizon View

by QNAP Systems Inc.


I’d like to welcome QNAP to the Login VSI Reference Architecture library. This is their first publication and has earned the “Tested with Login VSI” logo. Based in Taipei, Taiwan, QNAP offers a wide range of network attached appliances from smaller systems for home use all the way up to larger enterprise systems. In this case they are showcasing their ES1640dc Enterprise Storage product. It’s got what you’d expect… lots of SSD, data management features like dedupe, and integration with vSphere.


  • Users: 500 Knowledge Workers
  • Provisioning method: Linked clone
  • Provisioning system: VMware Horizon 6
  • Desktop OS: Windows 10, Office 2010
  • Hypervisor: VMware vSphere 6.0 Update 2
  • Hardware: Intel -based CPU E5-2620 v3 @ 2.4 GHz, ES and EJ shelves provisioned as iSCSI LUNs.

The cool bits

The performance of this system looked good and there was a particularly interesting bit of information that I enjoy seeing in these reference architectures which helps to highlight the storage system performance. As I’ve mentioned earlier, I believe write performance on storage is a critical operation that should be given the most scrutiny, and below QNAP shows us what they saw with respect to read/write ratios. Keep in mind that write operations usually carry a heavier impact on the storage system as it must perform additional operations for data protection, like the mirrored writes of RAID 10. While thinking about this also look at the high write IO percentage of standard run time during the Login VSI workload.

vdi reference architecture monthly november 2016 vdi production cycle for desktop infrastructure


This was a fun opportunity to learn just how much great information is contained in these documents. Please understand that the information that I found exciting is only part of the story and I highly encourage you to dig into anything that may catch your attention.

About this Publication

We are proud of the fact that technology vendors use Login VSI software to test their VDI solutions on a regular basis. Often, this testing results in a Reference Architecture with Login VSI test results. In this monthly newsletter, we review new reference architecture whitepapers published by technology vendors. Many of these documents can be long and time-consuming to review, so in this newsletter we try to save you time by distilling the highlights for you. Please note that the views expressed in this newsletter are solely those of Login VSI, and this newsletter is no substitute for a more thorough review of each reference architecture document on your own. We hope you find this newsletter useful!


Leave a Reply

Your email address will not be published. Required fields are marked *