Thin Clients Booting over a Wireless Bridge

How quickly can thin clients boot over a wireless bridge, and how far apart can they really be?

In the 1970s and 1980s, the ubiquitous model of corporate and academic computing was that of many users logging in remotely to a single server to use a sliver of its precious processing time. With the cost of semiconductors holding fast to Moore's Law in the subsequent decades, however, the next advances in computing saw desktop computing become the standard as it became more affordable.

Although the technology behind thin clients is not revolutionary, their popularity has been on the increase recently. For many institutions that rely on older, donated hardware, thin-client networks are the only feasible way to provide users with access to relatively new software. Their use also has flourished in the corporate context. Thin-client networks provide cost-savings, ease network administration and pose fewer security implications when the time comes to dispose of them. Several computer manufacturers have leaped to stake their claim on this expanding market: Dell and HP Compaq, among others, now offer thin-client solutions to business clients.

And, of course, thin clients have a large following of hobbyists and enthusiasts, who have used their size and flexibility to great effect in countless home-brew projects. Software projects, such as the Etherboot Project and the Linux Terminal Server Project, have large and active communities and provide excellent support to those looking to experiment with diskless workstations.

Connecting the thin clients to a server always has been done using Ethernet; however, things are changing. Wireless technologies, such as Wi-Fi (IEEE 802.11), have evolved tremendously and now can start to provide an alternative means of connecting clients to servers. Furthermore, wireless equipment enjoys world-wide acceptance, and compatible products are readily available and very cheap.

In this article, we give a short description of the setup of a thin-client network, as well as some of the tools we found to be useful in its operation and administration. We also describe a test scenario we set up, involving a thin-client network that spanned a wireless bridge.

What Is a Thin Client?

A thin client is a computer with no local hard drive, which loads its operating system at boot time from a boot server. It is designed to process data independently, but relies solely on its server for administration, applications and non-volatile storage.

Figure 1. LTSP Traffic Profile and Boot Sequence

Following the client's BIOS sequence, most machines with network-boot capability will initiate a Preboot EXecution Environment (PXE), which will pass system control to the local network adapter. Figure 1 illustrates the traffic profile of the boot process and the various different stages, which are numbered 1 to 5. The network card broadcasts a DHCPDISCOVER packet with special flags set, indicating that the sender is trying to locate a valid boot server. A local PXE server will reply with a list of valid boot servers. The client then chooses a server, requests the name of the Linux kernel file from the server and initiates its transfer using Trivial File Transfer Protocol (TFTP; stage 1). The client then loads and executes the Linux kernel as normal (stage 2). A custom init program is then run, which searches for a network card and uses DHCP to identify itself on the network. Using Sun Microsystems' Network File System (NFS), the thin client then mounts a directory tree located on the PXE server as its own root filesystem (stage 3). Once the client has a non-volatile root filesystem, it continues to load the rest of its operating system environment (stage 4)—for example, it can mount a local filesystem and create a ramdisk to store local copies of temporary files. The fifth stage in the boot process is the initiation of the X Window System. This transfers the keystrokes from the thin client to the server to be processed. The server in return sends the graphical output to be displayed by the user interface system (usually KDE or GNOME) on the thin client.

The X Display Manager Control Protocol (XDMCP) provides a layer of abstraction between the hardware in a system and the output shown to the user. This allows the user to be physically removed from the hardware by, in this case, a Local Area Network. When the X Window System is run on the thin client, it contacts the PXE server. This means the user logs in to the thin client to get a session on the server.

In conventional fat-client environments, if a client opens a large file from a network server, it must be transferred to the client over the network. If the client saves the file, the file must be again transmitted over the network. In the case of wireless networks, where bandwidth is limited, fat client networks are highly inefficient. On the other hand, with a thin-client network, if the user modifies the large file, only mouse movement, keystrokes and screen updates are transmitted to and from the thin client. This is a highly efficient means, and other examples, such as ICA or NX, can consume as little as 5kbps bandwidth. This level of traffic is suitable for transmitting over wireless links.


Geek Guide
The DevOps Toolbox

Tools and Technologies for Scale and Reliability
by Linux Journal Editor Bill Childers

Get your free copy today

Sponsored by IBM

Upcoming Webinar
8 Signs You're Beyond Cron

Scheduling Crontabs With an Enterprise Scheduler
11am CDT, April 29th
Moderated by Linux Journal Contributor Mike Diehl

Sign up now

Sponsored by Skybot