Last week there was lots of attention to my article entitled “Open letter to non-persistent VDI fanboys…”. There was a wave of comments. Twitter discussions went frantic, and the LinkedIn Desktop Superhero discussion board was the stage for a continued discussion on the subject.
After reading the comments and the heated discussions I realized I should introduce you to a new type of desktop pool management.
I like to call it Transient Desktop.
The transient model address application and data persistency desired by users, while preserving benefits and cost reduction from non-persistent deployment models.
First things first…
The persistent desktop – the simplest way to describe persistent desktops is to compare them to personal laptops, where users are able to customize the look & feel and deploy their own applications. Those applications and settings will always be there when the user logs back on.
I am an advocate for this type of desktop deployment when used in conjunction with newer storage technologies that allow for increased performance and reduced capacity requirements. (Learn more at Open letter to non-persistent VDI fanboys…)
With the persistent desktop model IT can still use the same tools they are familiar with to manage Windows environment without incurring in additional learning, licensing costs, or having to manage VDI in a different way than physical devices.
The non-persistent desktop – this is a special category of VDI, where desktops are refreshed at some point in time, usually after user log off, returning to its pristine state. Fans of non-persistent desktops claim this is the best way to maintain a fresh and clean desktop. I was a big fan of non-persistent desktop and wrote multiple articles, including this one (Floating Pools are the way to go….).
Since desktops are constantly refreshed, applications deployed by users or documents not saved at the right folder are not preserved across sessions.
The non-persistent model setback is related to the tools utilized to create the persistent desktop user’s experience. A number of different tools, such as namespace or app virtualization solutions, are needed to deliver this experience. That turns into cost, and ultimately into an environment with a hodge-podge of tools that IT needs to be familiar with in order to manage virtual and physical desktops.
Additionally, application virtualization is not the answer to all types of applications; and in large organizations with thousands of application this could easily become a hard problem to solve.
The cost saving benefits in the non-persistent model comes from reduced hardware infrastructure requirements, especially storage.
So we get to the new model.
The transient desktop – Imagine a world where we could have user data and application persistency along with benefits and cost reductions provided by non-persistent desktops infrastructure requirements, and still be able to manage VDI with the same tools IT is familiar with.
The non-persistent model cost savings comes from fewer infrastructure requirements, especially storage. Since desktops are refreshed upon user’s logoff there are no requirements for storage areas network (SAN). Desktops can simply exist in server-based SSD, or in RAM if you adopt a solution like Atlantis.
The non-persistent model setback is user data and application persistency. If a desktop or a host goes down IT is not able to re-instantiate the user’s desktop to the previous known state it was before the outage.
On the other hand, non-persistent server-based SSD solutions are cost-effective because it’s assumed desktops will be refreshed to its pristine state, saving on SSD storage capacity. That’s why DaaS providers like this model, the cost reduction due to no-SAN approach. However, SSD MLC costs are quickly coming down and today is possible to buy consumer grade 512GB SSD for as low as US$300.
Let’s say we are designing a solution with server-based SSD.
At the backend and monitoring desktops, both physical and virtual, is Horizon Mirage. Mirage keeps a centralized and de-duplicated copy of desktops, including user’s applications and data, and is able to re-instantiate them on different hosts should you have a host failure.
IT can still use the same tools they are familiar with to manage the entire Windows environment, virtual and physical, and will only need to interact with Mirage GUI should a host or desktop failure occur. In this scenario, using consumer grade MLC is not a big issue as desktops can be re-instantiate to its last known state in a different host.
Mirage can also distribute individual and departmental application layers, effectively replacing the tools IT is familiar with, for both physical and virtual. However, I am delving into this topic because this article’s subject is covering the new Transient Desktop VDI model, not application delivery.
In saying that, adding Mirage to the VDI solution IT is effectively removing the need for complex namespace or application virtualization solutions.
As of today, Mirage is not supported with Horizon View environments (view this post from Gabe Knuth). However, when VMware finally untangle the issues it will be possible to have a whole new breed of desktops that are persistent while they last.
How is that different than non-persistent desktops using Mirage? Non-persistent assumes that desktops are refreshed every so often. Transient desktops are not refreshed and exist in a no-San environment; drastically reducing implementation costs. But if for any reason you lose the desktop, it can be easily re-instantiated in a new host.
VMware does not currently support the steps outlined below. I recommend testing in development environment. If you decide to test or implement you are doing it on your own risk.
Some of my blog posts offer unsupported approaches for common problems and requirements faced by VMware customers. On the same spirit I would encourage you to deploy Horizon Mirage and start managing few virtual desktops using the Transient Desktop model.
Disclaimer: Any views or opinions expressed in this article are my own, not my employer’s. The content published here is not read, reviewed, or approved by VMware and does not necessarily represent or reflect the views or opinions of VMware or any of its divisions, subsidiaries, or business partners.
This article was first published by Andre Leibovici (@andreleibovici) at myvirtualcloud.net.