A Raspberry Pi Thin Client Computing Environment

Design, Configuration, Installation, and Testing

About this Guide

This guide was created to organize and simplify information about using the tiny, low-cost, energy-efficient computing platform called the Raspberry Pi or RPi as a thin-client. It provides guidelines for procuring, configuring, testing, and deploying a thin-client network using the RPi.

This guide is intended for computer-literate readers interested in the Raspberry Pi, computer virtualization, or computing energy efficiency. If you're not sure whether this guide is for you, that's OK. Have a look at the at the sidebar first to get an overview of the RPi and the topic of virtualization.

About the Raspberry Pi

About Virtualization

What is Virtualization?

Typical work environments still rely on traditional workstation/server computing environments in which files are stored on a company server and each worker uses a workstation to access and change those files.

In this scenario, the workstations do the computing, and the file server acts as a storage receptacle for files. This structure is inefficient, placing power-hungry workstations at every desk and creating a high degree of unused computing overhead on the server side.

Virtualization refers to the process of abstracting and presenting a virtual representation of computer hardware.

In our thin-client environment, this is done in order to allow a centralized server (called a hypervisor) to simultaneously divide its heavy computing power between multiple external, low-power thin clients, each of which interacts with the virtual hardware allocated to it.

This structure allows the server to run all of the computing tasks, consuming previously unused overhead, and allows the client to maintain a passive, low-power presence.

This also centralizes and organizes the computing needs of an organization, allowing for accurate scaling according to actual rather than projected needs. It is a very efficient way to provide computing resources to an organization

diagram of power-efficient rpitc environment

Why RPiTC?

The Raspberry Pi leverages the low cost and low power consumption of ARM-processor-based mobile phone platforms and reconfigures them into a small-footprint, general-purpose computer appropriate for experimentation and prototyping. Until now, the problem has been that although the Pi is powerful, it cannot efficiently run fully-featured operating systems with memory-hungry GUIs. The Raspberry Pi Thin-Client Project's solution is to use the Raspberry Pi as a thin client, offloading the computing power to a hypervisor.

for the remainder of this discussion and throughout this document, we will refer to the Raspberry Pi as the RPi and the Raspberry Pi Thin-Client Project as RPiTC.

Because the RPi is so inexpensive and RpiTC is free and open-source, the savings over even the least expensive commercially-available thin clients are tremendous. For large enterprises buying in volume, the cost of the RPi is low enough to be considered disposable. For users who need smaller scale deployments, RPis represent the most cost-effective way to move to a thin-client computing environment.

Why KVM?

Currently, the hypervisor market is dominated by only a handful of commercial offerings. Xen and KVM, by contrast, are open-source hypervisors that are nearly as fully-featured as commercial equivalents offered by VMWare and Microsoft. For organizations running critical systems, enterprise-level support contracts are available from Citrix, Xen's parent company. For those who do not need support contracts, the product is free. KVM, by contrast, does not offer support contracts and is fully open source, developed by the Open Virtualization Alliance. For the limited scope of this project, we have chosen KVM.

① Planning your Deployment

Before we jump in and start connecting and configuring devices, it's important to take a few minutes to think carefully about our planned setting for this deployment. This section takes you through the process of logically allocating existing resources and augmenting them with additional equipment and software where necessary. Once we are sure we have everything we need, we will make the initial equipment connections.

Assess your Needs

If you are creating a test environment at work and are concerned about appropriate computing resource allocation, you should take care to plan your deployment using this guide. If you are creating your thin-client network at home, this section will be less important.

Calculate Necessary Hardware Specifications

There are two main computing resources that need to be carefully allocated in a virtual environment: RAM and processor performance.

Of the two, RAM is significantly more important, since the amount of available RAM determines the absolute limit of virtual hosts possible. As a general rule, you should acquire and install the most RAM you can afford on your server.

In our lab, we are working with a Thinkserver TS140, which has a maximum RAM capacity of 32GB. If we want each of our thin clients to use 2GB of RAM, and we want to reserve 8GB for the server to use, you can see that we can run a maximum of 4 thin clients. This is a very conservative allocation plan that will ensure that a large amount of overhead is available.

For an initial testing environment, this conservative estimate is fine. As you progress into final testing stages, you will begin optimizing resources and incrementally increasing the number of clients to achieve the best balance of client and server performance.

about RAM and 32-bit processors.

Plan your Network Topography

The minimum hardware requirements for a thin client network:

Acquire your Equipment

You should acquire a server that meets your initial testing needs, with 2GB of available RAM per client. Note that this number is for initial testing purposes and the actual number of supported clients will increase during the final testing phase.

You should acquire the minimum number of RPis you will need for initial testing, with the understanding that the number of RPis in use will increase during final testing.

For our testing environment, we are using a Centos 7/KVM-based Thinkserver TS140 with a quad-core 3.2 GHz Xeon CPU and 16 GB of RAM, accompanied by one RPi.

For instructions on configuring KVM on CentOS 7, follow this guide.

The minimum required equipment list is:

1 server (64-bit processor, minimum of 8GB RAM)
1 RPi with power supply and case
1 formatted microSD card
1 staging computer running Windows Vista, 7, 8, or 10
USB keyboard
USB mouse
Computer monitor with HDMI output
HDMI cable

The following utilities are also needed:

7-Zip
Win32imager

Stage your Equipment

Follow these steps:

Power on your staging computer and open a web browser
Insert the formatted microSD card into the USB adapter, then insert the USB adapter into an available USB port on your workstation or laptop. Windows machines should mount the USB drive automatically.
Download Win32diskimager. When prompted, select Save.
Install Win32DiskImager. This will create a program group called Image Writer.
You must right-click the program launcher and select "Run as Administrator" to run this program.
Download and install 7_Zip.
When prompted, click Save File, then install 7-Zip.
Attach the USB keyboard and mouse to your RPi.
Attach a computer monitor to the RPi using the HDMI cable.

Do not power on the RPi yet.

② RPiTC Installation

In this section, we cover the process of installing and configuring the RPiTC operating system on the RPi. We conclude by connecting to our hypervisor over the network.

Download

Using the browser you opened on your staging computer, download the RPiTC operating system.
When prompted to open or save the RAR file, select Save.

Extract

Extract the RAR file you just downloaded using 7-Zip.
Select the path to the .rar file, then select an output path for the extracted files. Here, we are extracting to the Downloads folder.

Install

Install the PRiTC2 operating system using Win32DiskImager.
Locate the path to the file you extracted in the previous step, then install the image on the MicroSD card.

See this useful information regarding extraction and installation of images to RPis.

Boot

Unmount or eject your MicroUSB adapter and unplug it from your staging computer.
Boot your RPiTC by inserting the power adapter.

The RPi does not have a power switch and must be powered off and on by attaching and removing the power supply. This does not harm the RPi. Power switches for the RPi are available, but they are unnecessary for testing purposes.

Configure

Once you have booted the RPiTC operating system, you will see the default desktop.

The RPiTC OS comes prepackaged with all the remote connection utilities and scripts users might need to connect to their hypervisors.

To access configuration settings and automatically start a remotely-hosted virtual desktop on startup:

Select Session and Startup from the Settings menu.
Populate the connection client with the username, password, and port settings you configured when creating your virtual machines on the server.
Select Autostart for the type of connection you configured in your KVM server. This will automatically start the client and connect to your server as soon as your RPi finishes booting.

③ Testing

In order to gauge the performance and reliability of our RPiTC environment under heavy load, this section guides you through the process of testing, adjusting, and retesting your client and server.

Thin Client Testing

If using multiple clients, begin by connecting all clients concurrently using the RAM allocated during the initial steps of this guide.
Determine whether the CPU and RAM of the server can handle significant loads by playing back a remote video file of known bitrate and resolution through each client. This will test the server’s ability to continuously buffer and relay video to each client, a task which is computationally intensive, and which uses significant amounts of RAM.
If there are no performance problems, re-allocate RAM for each client by removing 512MB. Remember that we originally configured our clients to run on 2GB of RAM, so they will now be using 1.5GB of RAM. Retest the clients using the above method.
If there are still no noticeable performance issues, reduce the amount of RAM by a further 256MB, testing after each reduction. Once performance issues arise, add 256MB and re-allocate. This will be the final amount of RAM used in deployment.

Server Testing

On the server side, we are beginning with 8GB of RAM. We want to determine whether heavy client loads will degrade the performance of the server and reduce its ability to operate and allow for human interaction.

To do this, we will simulate several typical administration tasks to run concurrently with the client testing method described above.

Begin by running the video playback test for all attached clients, as above.
Log in to your server and run some common administrative tasks, such as backup and recovery, compiling a custom kernel, patching and updating, and disk health checks.

If these tasks cannot be carried out efficiently while clients are placing the server under heavy load, we have options for increasing that performance. These options, in increasing order of cost and complexity, are as follows:

Increase server RAM allocation and decrease number of clients.
Increase server RAM allocation and decrease client RAM allocation.
Increase server RAM by installing additional RAM.
Increase server RAM by installing additional RAM and increase disk access times by changing hard drive type or configuration and/or adding RAID capability.

It is up to you to determine which of these methods is most suitable for you based on your budget.

④ Next Steps

Virtualization offers several advantages over traditional computing environments. It is inexpensive, power efficient, avoids unnecessary computational redundancy, and is scalable.

The RPiTC project extends this logic, exploiting the availability of very inexpensive hardware and free, open-source software. Installing and configuring RPiTC requires simple, easily-available equipment and provides users with a simple method of connecting to their hypervisors.

This guide has provided the foundation for exploring the Raspberry Pi, RPiTC, and virtualization. Use it as a starting point for exploring your own hypervisor and thin-client solutions.

To hone your skills, try using what you have learned in this guide with a different hypervisor like VSphere.