One thing to consider when setting up ThinLinc is what type of hardware you require, and how much of it. These requirements are unique to each individual use case, and it is difficult to give any specific recommendations. However, there are some general guidelines which may help with server sizing for your project.
The ThinLinc server software is available for both 32- and 64-bit x86 architectures.
The ThinLinc server software has very modest hardware requirements. Most of the load on a ThinLinc server will be produced by the underlying operating system and user applications which run on top of it, not by ThinLinc itself. In most cases, the load produced by ThinLinc is negligable enough to be safely ignored.
An exception to this is where a large number users are expected to log in at the same time, for example in a classroom where lessons take place according to a schedule, or a workplace where everybody tends to arrive at once. In these examples, the ThinLinc master server will experience load spikes at the beginning of each lesson or workday, and will need to be sized accordingly.
What is the difference between a "master" and "agent"?
-
The ThinLinc master (
vsmserver) handles incoming connections, authenticates the user, selects the appropriate agent on which to create a new session, and keeps track of existing sessions. A ThinLinc installation has exactly one master. -
The ThinLinc agent (
vsmagent) is where user sessions are hosted and run. A ThinLinc installation may consist of one or more agents.
ThinLinc may be installed on a single server, or on multiple servers in a cluster configuration. When installed on a single server, the ThinLinc master and agent reside on the same machine.
In cases where login load spikes are expected, it is advisable to have a dedicated master rather than installing ThinLinc on a single server. This prevents the load spikes from happening on the same machine as user sessions, which may impact performance.
Since a ThinLinc cluster consists of exactly one master, you may want to set up a failover server in case the original master becomes unavailable. This is known as a “high availability” (HA) configuration. Support for HA is a built-in feature of ThinLinc, and will require additional hardware.
It is not necessary to over-provision hardware when sizing agents. Adding more resources to a ThinLinc cluster is a simple matter of adding more agents, which is a trivial task to perform, even on a live system.
ThinLinc will run equally well on both virtual and physical machines. Virtual environments may simplify the provisioning and management of ThinLinc systems, and certain platforms may provide load-balancing, redundancy and high-availability features which can be used instead of, or in addition to, those provided by ThinLinc.
For server-side graphic acceleration, a GPU is required. Sessions running on a physical machine with direct GPU access tend to give better performance than a virtual machine with GPU-passthrough, however results may vary.
In all cases, the best method for determining hardware requirements is to set up a proof-of-concept installation, and monitor resource usage for a small number of typical users over a period of time. This also gives end-users the opportunity to provide feedback on their experience. Since resource usage does not always scale linearly with the number of users, it is important to perform this testing with several users instead of just one. ThinLinc ships with 5 free licenses for this purpose. If you require more licenses than this for your testing, then please get in touch.
For more information on the ThinLinc server’s hardware requirements, see TAG chapter 3.2. Server Requirements
For more information on HA, see TAG Chapter 6. High Availability (HA)
For more information on using a GPU with ThinLinc, see TAG chapter 3.7. VirtualGL
ThinLinc can be downloaded and used free for up to 5 concurrent users from https://cendio.com/thinlinc/download