Last month, I joined Microsoft Cloud Solution Architect Jeff Mitchell to present the lessons learned deploying Citrix technologies on Microsoft Azure. In case you missed it, the webinar recording is available on our Technology in Practice homepage.

The event included a live Q&A portion at the conclusion of the webinar, but Jeff and I ended up with more questions than we could get to. We aggregated these questions here to address the key areas of interest that we couldn’t cover during the session… enjoy!

Q: Is Citrix Cloud the only way to host workloads on Azure? How else can I use Citrix and Azure?

Using the Citrix Cloud XenApp and XenDesktop Service is one of the many ways to deploy application and desktop workloads on Azure. While this was the primary method highlighted in this webinar, there are multiple ways a Citrix customer can get started using Azure.

Starting with 7.11, organizations can extend on-premises workloads using the Azure Resource Manager host connection in Citrix Studio. This involves configuring a secondary zone within an Azure datacenter and managing the workloads from a primary zone on premises. With this approach, customers need to be mindful of the latency between their primary zone and Azure. XenApp and XenDesktop 7.x sites can also be hosted within a customer’s Azure subscription. This would involve building a net new site in parallel to an existing on premises deployment. Each site would need to be managed separately with configurations replicated as required.

A lot has changed since we first started Azure support therefore it is recommended to use 7.15 LTSR or later with these deployment methods to take advantage of the latest Citrix features and Azure Resource Manager enhancements. Additionally, we covered these architectures in greater detail in our previous TIPs: Designing and Deploying on Public Clouds webinar. The on-demand recording is available from the following link.

Q: Are Availability Sets required for a Citrix Cloud XenApp and XenDesktop Service deployment on Azure? How would this differ from a traditional XenApp and XenDesktop deployment?

Datacenter availability is facilitated using Azure Availability Sets. Availability Sets can be used to separate Azure VMs into separate fault and update domains within a datacenter to reduce the risk of an outage during planned and unplanned maintenance. These are required to meet the Azure SLA for Virtual Machines of 99.95%. It is recommended that each tier of a Citrix deployment be in its own Availability Set. This applies for Citrix Cloud and a traditional 7.x deployment. For example, an Availability Set should be made for Delivery Controllers, another for NetScalers, StoreFront, etc. Eventually the ability to use Azure Availability Zones will release out of preview, further protecting customers from datacenter failures. Please note at the time of this blog the Azure Availability Zones feature is in preview only.

At scale, one key consideration of Availability Sets is the number of available fault domains within an Azure datacenter. Fault domains define the group of virtual machines that share a common power source and network switch. The number of fault domains depends on the datacenter, with up to three in most locations. When adding a 4th infrastructure component, for example 4 Delivery Controllers or Connectors to support 15k+ users/VDAs in N+1, there will be 2 components sharing a fault domain. Therefore, if a fault domain encounters an issue there is a chance for a larger than expected impact (i.e. 50% of components the instead of 25%).

For a large, business critical Citrix on Azure deployment, supplementing datacenter availability provided by Availability Sets with regional availability is a key part of business continuity planning. Regional availability can be accomplished by deploying an additional Resource Location in the appropriate Azure region pair, for example East US 2 and US Central Regions. Using NetScaler Global Server Load Balancing or Azure Traffic Manager can manage the routing requests between regions.

Q: What are the storage considerations for scaling out my XenApp or XenDesktop workloads?

There are two types of storage available in Azure, Standard and Premium. Standard Storage utilizes spinning disks and is meant for low cost, low criticality workloads. The challenge with Standard Storage for Citrix workloads is that IOPS and throughput of a disk is not provisioned, therefore there can be variance in the performance of a VM disk. Premium Storage utilizes solid-state drives for higher performance and low latency and requires instances such as DS-series, GS-series, Ls-series, and Fs-series. Performance is provisioned for VM disks providing more consistency for the workload.

Standard and Premium VM disks also come in managed and unmanaged varieties. Unmanaged disks leverage storage accounts. Storage accounts have a set amount of IOPS (20k for Standard, 50k for Premium) and these resources must be managed by the customer. From a cost perspective, there are charges associated with the IOPS and each unique GB consumed within a storage account. The disadvantage of storage accounts (unmanaged) is availability. Each storage account is within a storage scale unit (stamp). If a stamp fails due to hardware or software failure, the VM instances with disks on those stamps fail. When provisioning multiple storage accounts it is not possible to control the stamp where there accounts are stored.

With Managed Disks the storage resources are managed by Microsoft therefore the disks are allocated across multiple storage accounts to ensure they stay within scalability limits. Availability is also enhanced by including the stamp within the Availability Set of a VM, separating VMs across storage clusters. From a cost perspective, a managed disk has a cost based on the size of the disk, not the total GB consumed. Citrix infrastructure can be deployed using Managed Disks. Additionally, Managed Disk support is available for MCS workloads deployed with the Citrix Cloud XenApp and XenDesktop Service. At the time of this writing, support is not yet available for a traditional 7.x deployment hosted on Azure.

The table below can help you better understand the decisions associated with storage and Citrix workloads.

 EMPTY CELL BLANK Standard Premium
Managed Test/Dev environments.

Workloads where cost is more critical than performance.

Recommended for Production Citrix Infrastructure

Recommended for business critical workloads.

Unmanaged Test/Dev environments.

Not recommended for Production Citrix infrastructure.

Not recommended for business critical workloads.

Workloads where cost is more critical than availability.

Not recommended for Production Citrix infrastructure.

Not recommended for business critical workloads.

We also highlighted a few references during the webinar. They can be accessed from the following links:

  • Citrix on Azure Cost Calculator: A web based calculator that can be customized to help estimate the Azure network, storage, and compute costs of a Citrix deployment.
  • Citrix Education – Citrix XenApp and XenDesktop Service on Microsoft Azure: Our two and five-day courses cover the major aspects of installing, configuring and managing a highly available XenApp and XenDesktop environment on Microsoft Azure.

Citrix and Microsoft are here to help with if you have an upcoming cloud project. Simply submit the form here to speak with a Services specialist. Thanks very much for all those who attended and especially for you all who stayed on for the Q&A. We hope you found this information useful and you’ll consider joining us again in April for our follow-up event.  Subscribe to the Technology in Practice (TIPs) series here to receive notification once registration is live.