Azure cost saving by optimizing WVD workloads in Citrix Cloud.

Table of Contents

At GO-EUC, we’ve already written about the benefits of VDI optimizations from a performance perspective in the past. With the current buzz and increased interest in cloud workspaces such as WVD, are these optimizations still relevant and are there other benefits or caveats that need to be taken into consideration for those environments? With this research the focus is on the benefit from optimizing when using the new Windows 10 multi-session OS.

Windows 10 Enterprise multi-session

Windows 10 Enterprise multi-session (hereafter abbreviated to Windows 10 multi-session), is a new Remote Desktop Session Host that enables multiple concurrent interactive sessions, which previously only Windows Server could do. For Windows 10 multi-session, versions 1809 and later are supported and available in the Azure gallery. Please note that Windows 10 multi-session isn’t supported in on-premises production environments because it’s optimized for the Windows Virtual Desktop service for Azure.

Besides Windows virtual desktop entitlements, users running Citrix Virtual Apps and Desktops Service (Citrix Cloud) and VMware Horizon Cloud users are also eligible for Windows 10 multi-session.

Infrastructure and configuration

Windows 10 1909 multi-session is used as the default operating system from the Azure Gallery. Microsoft recommends to always use the latest version for performance and reliability. All required applications including, Microsoft Office 2016 x64 are installed using a post-OS installation task sequence in MDT. Citrix VDA version 1912.1 LTSR is used, as this is the most commonly used version at the time of writing.

The machines used are created using Citrix Virtual Apps and Desktops Service from the Citrix Cloud offering, running in a non-persistent scenario. MSCIO is not part of this research and therefore the MCSIO driver installation was omitted. To channel the communication between Citrix Cloud and the Azure resource location, two Cloud Connectors are deployed, using the Standard D2_v3 SKU in Azure. The on-premises GO-EUC AD is leveraged for authentication by using a site-to-site VPN.

One of the most used Azure SKU for VDI workloads is the Standard_D4_v2. This is an 8 vCPU machine with 28GB of memory. The Standard_D4_v2 comes equipped with a standard SSD as the premium storage is not supported on the D-Series. In this research the VM for the VDI workloads is equipped with a single data disk.

Dv2-series Azure VMs follow-on to the original D-series but have more powerful CPUs. The Dv2-series is about 35% faster than the D-series according to Microsoft. The Dv2-series VMs use the Intel Xeon 8171M 2.1 GHz (Skylake), the Intel Xeon E5-2673 v4 2.3 GHz (Broadwell) or the Intel Xeon E5-2673 v3 2.4 GHz (Haswell) processors.

According to the Azure Compute Unit (ACU) concept, the DV2 series has an ACU in the range of 210-250. More information on the ACU can be found here.

The estimated compute running costs for this VM are around €360 euros per month according to Microsoft own Azure estimates.


More information about VM sizes can be found here. Please be aware the listed VM sizes may not be available in your region.

For this research the following two scenarios are specified:

  • Standard_D4_v2 SKU, without optimizations, as the baseline;
  • Standard_D4_v2 SKU, optimized using Citrix Optimizer with the 1909 template.

Our testing methodology, as described here, applies to this research. Because of limited credits available in our Azure subscription, only 4 test runs are configured instead of the normal 10 runs.

More information on the various Citrix Optimizer optimizations can be found in another research here.

Expectations and results

Prior this research we did a poll on Twitter where we asked the community what they thought was the most important factor to take into consideration when choosing a new DaaS or Hybrid cloud VDI solution.