Failover solution for a terminal server?

Question

High availability

Good day!
The management set a non-trivial task for me to calculate the costs (without fanaticism) for transferring part of the office to thin clients (about 40 users). The solution must be fault-tolerant and scalable.
The question so far is about the scheme of work of services and the general principle of building a solution.
So far, two solutions are spinning in my head:
1) Terminal server
2) Server with virtual machines
So far I'm leaning towards the first solution, but I have no experience in implementing and maintaining either the first or the second. Therefore, I will try to state my idea and will be happy to hear from you that I am wrong :)
On three servers, the Remote Desktop Services and Remote Desktop Connection Broker service is raised, which in turn are combined into a cluster. We have a single entry point (virtual IP) where users connect and transfer them to one of the servers. In the event of a failure of one of the servers, the clients will reconnect to another server (unfortunately, not transparently. There will be a break in the session) Why three - so that there is a reserve of resources in the event of a failure of one node. In the case of an increase in resources, another server is introduced into the cluster and that's it.
User Folders and profiles - on two file servers with configured DFS service and replication between them. We have a single entry point and an exact copy of everything in the event of a failure of one of the servers (of course, without canceling the backup).
Even the following scheme was drawn:

In total we have 5 servers. A question about terminal licenses - do you need them for the number of users or will you have to purchase 40 of them for each server? And how is the idea in general? Is there anything with fewer resources?
The option with a virtual machine server seems to me a little less profitable in terms of resources and licenses. However, I'd love to hear any comments.
Once again I will voice the question - What are the schemes for fault-tolerant terminal solutions?
Thank you)

Answer 1

Have you considered the option of a terminal server in a virtual machine, for example, in a Hyper-v environment? There are good mechanisms in place to ensure fast failovers. But everything rests on the storage system, as in speed (which is completely solved either by clean SSDs or hybrid arrays using them (which is cheaper). But reliability remains in question. VDI turns out to be more expensive than a terminal solution, at least in my calculations :).
Replication of user profiles by means of DFS IMHO is not a good idea, people who tried to do this had problems. I would suggest ZFS and snapshots, but probably it will be past the checkout.
Well, the allowable downtime is also important. Even though management usually doesn't want to hear about any downtime (and sometimes it's justified), it needs to be realistically assessed.

Answer 2

In total we have 5 servers. A question about terminal licenses - do you need them for the number of users or will you have to purchase 40 of them for each server?

Terminal licenses must be purchased by the number of users. The number of servers does not matter. If we compare the costs head-on, then the capital costs for a VDI solution are higher than for a terminal one, because you will need to buy client Windows operating systems with Software Assurance, or an annual VDA subscription (depending on the type of client you are connecting from), which is clearly more expensive than one terminal license.

On three servers, the Remote Desktop Services and Remote Desktop Connection Broker service is raised

It is better to spread the terminal roles to different servers (even better - if these are VMs - it will facilitate the management of this entire zoo). Most likely, you will want to load balance between RDS Session Host servers using Microsoft NLB, while for Connection Broker, failover is provided by Microsoft Failover Cluster. Microsoft does not support NLB and Failover Cluster on the same server. The way out is to use an external balancer (Citrix, F5, Kemp, etc), or distribute roles across different servers.

User Folders and profiles - on two file servers with configured DFS service and replication between them.

Here are the scenarios where DFS is supported for the case with roaming user profiles: blogs.technet.com/b/askds/archive/2010/09/01/microsoft-s-support-statement-around-replicated-user-profile- data.aspx
As an alternative, use an external NAS server that supports the CIFS (SMB) protocol, or deploy another cluster of file servers on Windows Server 2012 with Cluster Shared Volumes, and shared reliable block storage (SHD).

Answer 3

omnimod , with a rough approximation, what is the cost of implementing the solution you described only due to software licenses, and this is without taking into account the licensing of applications for which such an architecture is actually built (what are these 40 users working with)?