Get the scoop on elastic application layers from Citrix
IT pros can use Citrix Elastic Layering to target apps to specific users. However, it's important to know when and where to use Elastic Layering, as well the potential problems.
Application layers allow IT to deliver applications to users without actually installing the apps on the virtual desktop operating system. The Elastic Layering feature in Citrix App Layering takes things a step further by allowing IT to deliver specific apps to certain users or groups of users.
IT professionals can use Elastic Layering to simplify image management because they don't have to add apps that only a few users work with to the desktop image. Instead, they can assign those apps to the users that need them based on factors such as Active Directory group membership or location. The apps are then available to users as icons on their desktops when they log on.
IT pros should be sure to only use Elastic Layering where it makes sense, though. If the app has a kernel driver, a third-party driver or a dependent service, for example, it won't work well elastically.
In addition, if thousands of users need a specific app, it's better to just put the app in the image rather than have all those connections go back to the file server. It's actually even better to just add another image in such cases, said Rob Zylowski, senior architect for Citrix Consulting Solutions, in a session at Citrix Synergy 2018.
How does Elastic Layering work?
Windows performs an in-guest mount -- an application program interface call to the shared mounts of the virtual hard disk (VHD) -- which then becomes a volume on the machine that the filter driver can see and handle.
Because the elastic application layers are on a share, none of the apps will work if the server fails. To mitigate the risk of failure, IT can use a server cluster to host the elastic application layers so that if one node fails, it will fail over to the next one.
Citrix Elastic Application Layering
If IT pros are going to invest in Elastic Layering, they should use some kind of cluster, according to Zylowski.
How does Elastic Layering handle scaling?
Scaling out elastic application layers is complicated because each app uses the VHDs differently. Some apps load everything when they start, whereas others talk back to the VHD constantly, which leads to high IOPS.
Test your knowledge on the fundamentals of app layering
Demonstrate how much you know about the kinds of criteria IT uses for app layering, where app layering hard disks live and more.
To handle these fluctuations, IT pros must use a pod architecture -- a collection of servers -- that can handle different numbers of shares because each machine has different registry settings that point to the elastic shares, Zylowski said.
A key concern with Elastic Layering is that it can add time to users' logons.
IT can use a Group Policy Object to define where each machine will look for elastic shares. Before deploying on a large scale -- say, 5,000 shares at once -- IT should test the deployment on a small scale of about 300 shares to see what the load is when users work with the elastic application layers. If that works, build up with another 300 shares and so on until IT hits its full deployment number, Zylowski said.
What are the benefits and drawbacks of Elastic Layering?
Elastic application layers are read-only, which helps performance; most file servers are better equipped to process reads than writes. The only caveat here is that VHDs are large, so it's important to optimize them for reads, Zylowski said.
A key concern with Elastic Layering is that it can add time to users' logons. With elastic layers, logon times come down to how long it takes the layers to mount. That's decided by two factors: how long it takes the Windows client to perform the mounting and how long it takes the file share to respond.
IT pros should not skimp on their file server with Elastic Layering or they will run into logon time issues because they don't have enough RAM to cache.
If IT pros use Citrix XenApp with Elastic Layering, they can set a machine object in a group and assign elastic layers to that machine object rather than to a user. This allows them to assign apps to a specific silo of machines. In these situations, they can use a synthetic logon to preload the elastic layers so that app logon is faster because, when a real user logs on, nothing has to load, Zylowski said.
