What's your VDI solution at work?

f1y

Supreme [H]ardness
Joined
Dec 30, 2005
Messages
8,107
We currently have an on-prem VMWare infrastructure. tons of expensive cisco blades, and storage. Tons of over-provisioned resources on win10 VMs.

Maybe 500 windows 10 VDIs, and citrix loaded on them for remote delivery. The users hit storefront, and launch their desktop.

I have a project of going with "VDI-in-the-cloud"

I'm curious what other people are using? What's your delivery to the end user like?

Horizion, aws workspaces, citrix on citrix cloud, tell your devs to just use hyperv locally?

What was some of your trouble migrating to that solution?
 
Horizon 7, give users a full Windows 10 desktop with basically the same image as our desktop one. We also have RDS and Citrix if they just need to access specific apps, RDS can be front-ended by Azure if you want to "cloud" it for remote access. Not sure I'd want all my apps running in the cloud if users are working on-prem anyway, what's the use case for your VDI in the cloud?
 
Use case is remote users. Offshore and consultants. Keeps people working for us on a standard issued image/build to reduce troubleshooting.
 
While on a smaller scale, we're currently using RDS for all of our VDI needs (Smaller scale = 100 or so Users).
 
If you use citrkx provisioning and do random vdi assignments with networked user profiles and slimmed down as needed profile copying you will saflve yourself on bandwidth. And citrix can work with your hypervisors to shut down not used workstations. Works pretty well and since cloud is based on resource consumption it will help.
 
I have a project of going with "VDI-in-the-cloud"

I have been working on a "VDI-in-the-cloud" project as well over the last several months. Using VMware VDIs. Also with an already existing VMware horizons VDI on-prem (However the existing VDI is only used by remote employees).

We colo'd vmware servers in a datacenter and have data replicating using DFS + PeerLink software.

The biggest issue/pushback we have been getting in our testing is from users who are used to working on a local machine. Just the minor latency/delay when dragging things around or scrolling through documents. The users are just not used to even a tiny amount of delay. Existing VDI/remote users have been happy.

I think going into it with the outlook that is a large shift in VDI strategy in mind is key. Being able to be more distributed, more remote, less infrastructure in house are the pluses. However switching users from a PC at their feet to a VDI has definitely caused a bit of a headache.
 
Easius, you need to get your latency down. If your users are seeing issues with use because of delay let your cloud providers know and see if they can start geo fencing your VDI's to improve latency. Users IP shows they are in Ohio, they get something in that region not South west Texas. And so on. Could solve your use latency issue or at least make it 'good enough' they don't have a problem with it.

If you don't do that yea it is cheaper over all but you can have some pretty high latency.

Also your users have to have a decent pipe as well. Have your network team check routes. (If you haven't.) There may be a more optimal route in your company to route to the cloud provider in question.

You could even do the geo fencing based on who the user is and where they will be logging into. And STILL keep the 'random vdi host' thing going.
 
Easius, you need to get your latency down. If your users are seeing issues with use because of delay let your cloud providers know and see if they can start geo fencing your VDI's to improve latency. Users IP shows they are in Ohio, they get something in that region not South west Texas. And so on. Could solve your use latency issue or at least make it 'good enough' they don't have a problem with it.

If you don't do that yea it is cheaper over all but you can have some pretty high latency.

Also your users have to have a decent pipe as well. Have your network team check routes. (If you haven't.) There may be a more optimal route in your company to route to the cloud provider in question.

You could even do the geo fencing based on who the user is and where they will be logging into. And STILL keep the 'random vdi host' thing going.

Main offices are around 4ms and 15ms to colo both on SLA'd fiber circuits of 100/100. We've tried shaping traffic in various ways as each office has multiple circuits but still get complaints.
 
Main offices are around 4ms and 15ms to colo both on SLA'd fiber circuits of 100/100. We've tried shaping traffic in various ways as each office has multiple circuits but still get complaints.

Is this something you can recreate if you bring your laptop in?

We had an issue with a bunch of workstations using web apps. Performance on our own in IT was fine. Figured out it was McAfee DLP scanning anything that came into or went out of the problem PC's. White listed the site and wala problem was solved. Took me 1 day to solve it 3 days to prove it. Sounds to me like you have something injecting latency into the input to your VDI's and I am betting it originates from the client.
 
We currently have an on-prem VMWare infrastructure. tons of expensive cisco blades, and storage. Tons of over-provisioned resources on win10 VMs.

Maybe 500 windows 10 VDIs, and citrix loaded on them for remote delivery. The users hit storefront, and launch their desktop.

I have a project of going with "VDI-in-the-cloud"

I'm curious what other people are using? What's your delivery to the end user like?

Horizion, aws workspaces, citrix on citrix cloud, tell your devs to just use hyperv locally?

What was some of your trouble migrating to that solution?


2000-3000 (~6000 peak) users, local and remote still on Win7, Win10 expected this year on Nutanix. We are exploring the same things. Everything from Horizon DAAS to AWS workspaces or Xi Frame. Some remote sites average 250ms latency. If we are able to push geofencing, it should improve experience drastically. We tried going DAAS last year but for some reason, getting the 10g circuits was going to take too long so the project was scrapped and now starting over.
 
Last edited:
Since you are using Citrix StoreFront to connect to VDI (via Receiver), you can install latest version of Citrix Studio (7.18 should be the one), but make sure, when it domes to XenAPP/XenDesktop on VDI, you use either version 7.15 or 7.14. DO NOT use 7.18 version of XenApp/Deskop, it has known issue of slowness (rendering on Receiver) and many printer redirection related problems.
 
We currently have an on-prem VMWare infrastructure. tons of expensive cisco blades, and storage. Tons of over-provisioned resources on win10 VMs.

Maybe 500 windows 10 VDIs, and citrix loaded on them for remote delivery. The users hit storefront, and launch their desktop.

I have a project of going with "VDI-in-the-cloud"

I'm curious what other people are using? What's your delivery to the end user like?

Horizion, aws workspaces, citrix on citrix cloud, tell your devs to just use hyperv locally?

What was some of your trouble migrating to that solution?

Workspaces, but the Cloud 9 ide was the thing that made me happy. Secrets management/injection is safer. Managing AD is a breeze. Rescinding role access upon "sec event" is a breeze.

There's less time burned administrating dog food and more time to do more meaningful things.

My issues when I was primarily a VMware guy was the janitorial. A good NA you can partner with made or broke my day. Workspaces is easier bc I'm the NA for the most part. Direct Connect is something I can manage. Additional Workspaces VPCs get iterated off my Cloudformation template. QoS issues are encapsulated in the specific VPCs requirements.

ie if you like Python, JSON, and yaml then getting in the weeds remediating QoS is just iterating Lambda functions that trigger off tailored Cloudwatch Events. It feels like "automation-y".
 
Last edited:
Simple Windows 10 Multi session via Azure and VMs hosted inside our tenet using FX Profiles. Also a Remote App server as well.
 
Back
Top