Choosing the Provisioning Model for Image Management
One of the most common design decisions that needs to be done for every Citrix Virtual Apps and Desktops (CVAD) project is which provisioning model meets the business and operational requirements. The goal of this article is to describe the most common decision factors, recommendations, and different scenarios where certain provisioning model might be a better candidate. For image management, there are two provisioning models that are commonly used by Citrix administrators to manage their Citrix environment efficiently:
- Machine Creation Services (MCS)
- Citrix Provisioning (PVS)
It is also important to mention that Citrix App Layering is out of scope for the current version of this document. Implementation of Citrix App Layering can influence many of these design decisions and we will include it in one of the future updates for this article.
This article is focused on design decisions and factors involving image provisioning. If you are interested in more general reference architecture for Citrix PVS or MCS, we highly recommend reading Citrix Virtual Apps and Desktops Image Management reference architecture.
Overview of Citrix Provisioning Services (PVS)
Citrix Provisioning is a software-based streaming technology that can deliver centralized, shared operating system image to multiple virtual or physical endpoints. For design purposes, it is important to understand that PVS is active component - it is actively involved in the daily operations of image management and delivery. As an advantage, Citrix PVS can reduce the operational and storage costs, since it acts as a software-based storage offloading solution. This advantage, however means that the environment needs to be properly designed and maintained. Citrix PVS requires dedicated set of streaming servers, database, and needs to be included in high-availability planning. Design of the PVS environment is mostly hypervisor agnostic and implementations are similar for different hypervisors.
Overview of Citrix Machine Creation Services (MCS)
Citrix Machine Creation Services is an orchestration component of Citrix Virtual Apps and Desktops that can provide single image management for shared or dedicated machines. MCS is passive component - in most deployments, MCS is involved only in the image build orchestration process (telling the hypervisor what and where to do) and not in daily operations and delivery of images. There are some exceptions to this rule, most notably for hypervisors that cannot automatically reset disks. From a design perspective, environments built using MCS inherits the behavior and characteristics of the hypervisor or cloud provider that is hosting workloads. Design of the MCS environment is therefore heavily influenced by a combination of hypervisor and storage used.
Provisioning Decision Factors
Each project and environment is unique and have different requirements and goals. For that reason, it is common that a good architect chooses different provisioning models for different projects and not exclusive prefer only one. It is common that different provisioning models are used even in a same environment - for example when providing a combination of dedicated and shared machines.
For this document, we are going to divide decision factors in two categories - factors where it is clear which provisioning model is preferred (or have to be used) and factors that are more open to interpretation and where personal preference / experience is playing much bigger role in decision making.
Explicit Decision Factors
Explicit decision factors cover scenarios where PVS or MCS is not only preferred, but often there is only one possible option.
If your project involves provisioning to physical machines (typically classrooms or similar use cases), only Citrix PVS is supported (and functional). This model works well for standardized desktops such as those in lab and training environments, call centers, and “thin client” devices used to access virtual desktops.
Recommended model: PVS
If you are planning to run virtual apps and desktops in one of the public infrastructures as a service (IaaS) environments like Microsoft Azure, Amazon Web Services or Google Cloud Platform, be aware that such environments do not support Citrix PVS. There are technical limitations that prevent PVS from running in such environments - most notably the requirement for PXE firmware, which is required even when the BDM boot partition is used. While this limitation prevents you from running Citrix PVS in true cloud environments, it is still possible to run them in hosted virtualized environments.
Recommended model: MCS
When deploying virtual desktops in persistent mode, these are the most common approaches:
- Full clones with MCS
- Fast clones with MCS
- User layers
- Manual / ESD provisioning (SCCM)
- Hypervisor templates
While it is theoretically possible to provide dedicated desktops with Citrix PVS using private image mode, this approach is not recommended and does not provide any operational or performance benefits. Without the use of separate technology for persistent user layers, Citrix PVS is not a recommended model for persistent machines.
When using MCS for persistent machines, there are two possible approaches - using fast clones or full clones (introduced in version 7.11). While fast clones with MCS provide the benefit of small storage footprint and fast create and reset times (small delta disks), storage migration / backups / high-availability is more complicated in this deployment model. Since this is often a requirement for dedicated / persistent machines, full clones with MCS are the recommended approach to persistent desktops. You can read more about the difference between fast and full clones in article CTX224040.
Recommended model: MCS
Citrix PVS is not available to all license editions - most notably, you are not eligible to use it with Virtual Apps Standard (formerly XenApp Advanced) and Virtual Apps Advanced (formerly XenApp Enterprise) licenses. It is a common misunderstanding that it is possible to use lower editions of Citrix Virtual Apps with purchase of Provisioning Services data center Edition, however this is excluded in the license agreement. For more information, read Products and License Models.
Recommended model: MCS
Variable Decision Factors
While the previous sets of decision factors are straightforward, this second set is much more flexible, open to interpretation and personal preference / experience with technology plays a much bigger role in decision making. Citrix gives you flexibility to choose the best solution for your needs and your decisions for the following factors might be different than our recommendations.
This is an important factor to consider, especially if you are a Citrix partner and you are building a green field environment for a new customer. Consider the skills and capabilities of the team that is going to manage this environment - if a customer is new to Citrix technologies, is operating a static environment with minimum changes or if they have multiple roles and Citrix management is only a subset of their responsibilities, it might be a good idea to minimize the complexity of the environment and reduce the number of moving parts. In that case, MCS might be a better solution.
Recommended mode: MCS if technical skills are concern
Familiarity With Provisioning Model
Many of the Citrix customers and partners are familiar with Citrix PVS or MCS and have provisioned thousands of machines using this technology. This is the important factor when deciding which provisioning model to use - if you and the rest of your team is familiar with the procedures and technical aspects of one solution and that solution meets all your requirements and needs, use it for your new project. However, if you are a partner or build solution for third party, it is important to consider the learning curve and skill set of your customer.
Both PVS and MCS can support complex architectures and large environments - personal preference / experience is often the most important decision factor involved in choosing one.
Recommended mode: PVS or MCS, whichever model is more familiar
Complex Multi-Site Architecture
In certain type of environments, the ability to quickly replicate images across multiple storage repositories is critical. With PVS, this replication is simply and usually involves simple copy operation across different file shares. With enterprise MCS design, this requires replication of a master image itself together with automated image provisioning using CVAD SDK / PowerShell.
While it is possible to automate multi-site deployments with MCS, the PVS process is simpler and easier to use.
Recommended mode: PVS preferred
Requires Frequent Changes
One of the biggest advantages of Citrix PVS is the ability to almost instantly switch from one virtual disk (vDisk) to another and support for advanced versioning of virtual images. It is possible to achieve similar results with MCS using rolling catalogs and versioning on a master image level, however this process is simpler with PVS.
For environments that require frequent changes (multiple images changes every week), PVS might offer a more flexible, out of the box solution. There are more factors involved in this decision - for example how long does it take to update images using MCS in your environment and how many storage repository replications are required, but generally, you can expect PVS to be the more flexible image management solution.
Recommended mode: PVS preferred
Size of Environment
One factor that is not as important as many people believe is the scale of the target environment. Both PVS and MCS are enterprise ready solutions that can scale to tens of thousands of machines.
When scale is a potential decision factor is if you are designing a small and simplistic environment - unless there are some other factors involved (such as provisioning to physical machines), MCS is the preferred method for smaller environments (tens of machines).
Recommended mode: MCS for smaller environments, PVS/MCS for larger environments
Citrix PVS is sensitive to a properly working network environment - whether it’s proper routing/size of packets or stable network connection. Because it is using a hybrid UDP traffic, the impact of dropped packets can be substantial, as it requires a repeat of the whole sequence of packets. If network performance or stability is a concern, MCS (preferably not using NFS) might be a better approach.
Recommended mode: MCS if network stability is a concern
Requires Persistent Disk
There is a common requirement to keep some data persistent between reboots - for example event logs or configuration that needs to be restored after machine changes are deleted (for example unique machine identifiers used by anti-malware or software deployment tools to identify target machine).
With PVS and newer versions of MCS IO drivers (introduced in version 7.9), it is possible to store persistent data on write cache disk. This capability is possible with older versions of MCS, however it requires more scripting and automation skills. If you are not willing to automate this procedure or don’t have the required skills, using an out-of-box functionality might be a better option. Be careful when redirecting data to a write cache disk - not properly planning and monitoring free capacity can lead to stability issues and needs to be carefully considered before implementation. Best candidates for redirections are smaller files with fixed size (e.g. log files with maximum size or small text files), it is not recommended to redirect large or unpredictable amount of data.
Recommended mode: PVS or new version of MCS IO preferred
Optimized Hypervisor and Storage
As mentioned at the beginning, PVS is mostly a hypervisor agnostic solution, while the performance, stability and flexibility of MCS has a strong dependency on the underlying hypervisor and storage.
If your underlying infrastructure is however optimized and designed to work properly with MCS, it is possible to achieve better results with MCS, as you are going to use hardware-acceleration instead of software acceleration.
The most notable candidate to mention here is the Nutanix implementation of Shadow Clones, which is optimized for MCS provisioning. Another good examples are hypervisors that are optimized for virtual desktop workloads - for example Citrix Hypervisor with support for In-memory Read Caching or IntelliCache.
Recommended mode: MCS if using hypervisor/storage optimized for MCS
In this article, we have discussed the most common decision factors when choosing a provisioning method for your Citrix Virtual Apps and Desktops environment. Both Citrix PVS and MCS are enterprise-ready solutions that offer great performance and flexibility.
In this article
- Overview of Citrix Provisioning Services (PVS)
- Overview of Citrix Machine Creation Services (MCS)
- Provisioning Decision Factors
- Explicit Decision Factors
- Variable Decision Factors