The debate over the strengths and weaknesses of thin clients versus fat clients in a distributed computing environment has gone on for many years. Thin clients have been highlighted as a preferred method for information publishing across the enterprise and as a key tool in the ongoing struggle to reduce ownership costs for information technology. In late 2003, unsatisfied with its strategic direction and relieved of the most severe anti-trust threats, Microsoft began to reverse the technology pendulum back toward fat-client architectures as it announced strategic plans to embed more functionality within its Windows client operating system.
Overlooked in many discussions of industry trends is the “Zero Client“, a technology which offers the benefits of fat clients while delivering equivalent cost of ownership reductions and faster performance than the fastest thin clients.
The zero client (“station”) is a set of components (monitor, keyboards, mouse), none of which have independently programmable intelligence, that relies on a centralized CPU (“Host PC”) for all program execution and information processing. The connection between the zero client and the Host PC is a direct, point-to-point connection that operates at bus speed, requiring no network protocol. Zero clients are typically implemented in clusters, using a “star-like” configuration around the Host PC.
Each cluster can function either as a network component of a distributed computing system or as self-contained, small-group system. When combined with the high performance of the bus-speed delivery system, zero client technology offers an unequalled platform for small-group, transactional-based systems accessing a shared database. Since a zero client uses low-cost component hardware, with no local intelligence or processing, its cost per seat is similar to that of network computers. Likewise, zero clients offer a single point location – the Host PC – for upgrade, maintenance and support, thus drastically reducing licensing and lifetime system costs.
Zero client technology has its earliest roots in mini/mainframe computing, where computing tasks and program execution were centralized and information was sent and displayed to multiple users through terminal devices that lacked programmable intelligence, ergo, “dumb terminals” (later renamed “mainframe interactive terminals”).
Character-based terminals such as the initial 3270, 5250 and VT52/VT100 stations provided the user interface on a variety of systems. These terminals were typically connected to the host via low bandwidth serial links (i.e. less than 9.6 Kbps). Output from an application program was passed by the operating system through the serial link to the terminal firmware to be displayed on the user’s screen.
When personal computers were introduced, their computing architecture was a radical change for the industry. In the PC, applications could be executed locally on the user’s desktop, eliminating the requirement that the operating system transmit the output to a slow, external display device. Some of the earliest PC applications were terminal emulators so that a single PC could displace the dumb terminal on the desktop.
The impact of this change in architecture was dramatic and rapid. Applications began to change as developers embraced the assumption of “one user, one PC”. Using this dedicated-user assumption, PC applications began leveraging direct access to the hardware for maximum performance. For example, the user interface was optimized by bypassing the operating system entirely and directly addressing the display device.
Then, in the mid-1990s, coincident with the improved performance in newer Intel x86 chipsets, the PC user interface shifted from character-based to graphical. Windows and OS/2 became the predominant operating systems for Intel-based personal computers. In these advanced environments, the operating system took more control of access to and use of the PC hardware. In display management, the operating system was reinserted between the application and the display adapter. As a consequence of the relentlessly-increasing operating system functionality and more complex applications, it became more difficult and more expensive to provide support for the PC environments.
During this same period, the zero client technology (still using a serial connection) delivered less and less perceived performance as a direct result of the increased amount of information being passed over that connection for increasingly graphical and “user friendly” applications. In comparing the less than 2,000 bytes in a character based screen to the just over 300,000 bytes to represent the graphic pixels in the smallest Windows display, it became evident that the serial connection no longer provided a viable solution.
A new type of connectivity hardware was introduced in the late 1980s, generically referred to as a multi-display adaptor (“MDA”). These add-on boards contained multiple VGA chipsets and used a variety of cabling options (fiber optic, coaxial, etc.). When adapted to an operating system environment, they all delivered the display data directly to multiple VGA displays at bus speed. During the early to mid 1990s, these multi-display adapters were implemented for use on a variety of flavors of Unix (including SCO), other proprietary operating systems (including PC-MOS, VM386, THEOS) and enhanced DOS operating systems (Concurrent and Multiuser).
These MDAs were the early predecessors of the hardware used by zero client technology today. Today, multi-display hardware uses SVGA/XGA chipsets, supports 1600×1200 resolution in full color and directly delivers the video streams to multiple displays via a variety of high speed transmission media.
The use of this hardware with its bus transfer speed for additional video displays provided the hardware foundation necessary to deliver efficient zero client technology for Windows operating systems.
A traditional PC has a single display adapter, a single mouse port and a single keyboard controller. A zero client PC Host has multiple display adapters, multiple mouse ports and multiple keyboard controllers. Through system software that resides on the PC Host, multiple virtual machines or sessions are created, each associated with a display adapter, a mouse and a keyboard. Input for the session is read directly from its mouse and keyboard; output is written directly to its display adapter. As in any computing architecture, there are both hardware and software components involved to deliver this advanced functionality.
Key Components of Zero Client Technology
• Host PC – Standard PC with multiple display adapters
• Local Station – Standard (or USB) input and output devices(e.g. monitor, mouse, keyboard, audio, Touch Screen, serial, etc.)
• High Speed Delivery System – Direct connection or extension
• Software Component – Multiuser or Virtualization Software
The Host PC has more than one display adapter (or possibly a display adapter with multiple SVGA chipsets) for support of zero client technology. Each of those SVGA chipsets is associated with a direct connection to a local station, typically from 5 to 500 feet away. One example of a local station consists of a connector box of some kind, into which a monitor, mouse and keyboard are plugged. If local peripherals are used, the connector box will also include signal decoding, which will multiplex the combined video, serial data and parallel data and feed it to and receive it from the appropriate component.
A vital element of the zero client solution is the ability to transmit a true graphical signal directly to the station’s display, perhaps as much as several hundred feet away. By extending the VGA signal, as opposed to packetizing the video with software and network protocols, the Host PC is not burdened with CPU overhead and the responsiveness of the station’s display is as fast as a standalone PC system.
As indicated earlier, zero client technology has existed in various flavors for many years, However, until the introduction of the Applica software in late 1996, zero client technology had never been implemented on a Windows 95/98 platform.
Windows 95/98 included preemptive multitasking capabilities and was the first Windows-based platform in which zero client technology could be effectively implemented. In prior versions of Windows (3.x), multiple applications could be open in their own windows, but only one application was active at any given point in time. For example, a user could have Word and Excel windows displayed, but, after beginning a long recalculation in Excel, the user couldn’t switch to Word until the Excel computation was complete.
Using the above example with the preemptive multitasking in Windows 95/98, a user could have both Word and Excel open, start a long recalculation in Excel and then immediately switch to Word to edit a document while the recalculation finished in the background.
At a lower level, a Host PC using zero client technology has system software enhancements that support multiple virtual machines or sessions. Each of these sessions is associated with a display adapter, a mouse, a keyboard and optional audio. As previously mentioned, he system software directly passes input for each virtual machine from its corresponding mouse and keyboard; similarly, output is written directly to the corresponding display adapter.
One of the obvious benefits of the zero client design is very high video performance. The physical presence of a video chipset for each of the virtual machines eliminates the overhead of emulation, packetizing and transmission of graphical orders or video. The degree to which this benefits performance is directly tied to the extent that color and graphics are used by application(s) being executed in that virtual machine.
In addition, performance is improved because all display data is transferred at bus transfer speeds rather than through a network connection. A network connection requires the transmission of data in packets and using some protocol. The effective throughput of a network at any point in time is determined by multiple factors, including the bandwidth and amount of active traffic on the network at that time. The point-to-point transfer of display data directly from a memory structure to a video display can occur in a small fraction of the time required to pass the same data over the network.
Zero client technology also offers simplified installation, configuration and support, by virtue of the use of a single Host PC and multiple stations (each consisting of a monitor, mouse and keyboard) rather than multiple PCs individually configured and combined into a small network.
Thin client technology has received a lot of attention in recent years. In support of an industry focus upon expense reduction and improved manageability of desktop computing, the computer industry has drawn from the experience of the mini/mainframe model of host and terminal. With the thin client architecture, the application moved back to a multi-user host, which transmitted the display information to an intelligent device for presentation to the user.
However, the thin client model ignored a crucial change that occurred in the application domain with adoption of Windows as a standard platform: the move from a character-based to a graphical user interface. The client station must now do substantially more processing than the old “dumb” terminal. Higher bandwidth links are also required for the graphical information. When multimedia is added to the application equation, the effectiveness of thin client technology is severely reduced.
Zero client technology differs from thin client technology in client hardware requirements, display data processing and the data delivery system. A comparison of the processing of graphical commands points to some key differences. As a baseline, on a standalone PC, Windows passes graphical commands directly to a display driver that interprets them and updates the display.
Within a thin client host (terminal server), a protocol layer is introduced. Here, Windows passes graphical commands to a protocol layer, usually either the Citrix ICA or the Microsoft share (RDP). This protocol layer encodes the commands into packets and transmits them over the network to the intelligent client device. At the client end, the protocol layer decodes the commands and passes them to a display driver that interprets them and updates the display. Sun offers a similar capability with its Sun Ray line of products for the Solaris operating system.
On a zero client system, the process is almost identical to that occurring in the standalone PC, with the single exception that the driver updates the display that corresponds to each virtual machine or session. Within the zero client Host PC, the protocol layer and the transmission of the data in packets are avoided. Therefore, zero client architecture conserves processing power within the Host PC and eliminates client processing entirely.
The zero client architecture and combines the best attributes of the thin client and the standard personal computer architectures. As in the thin client, applications execute on a shared Host PC. This minimizes cost, delivers the highest performance and improves manageability.
The zero client architecture combines key aspects of the thin client, the NC and the personal computer. As in the thin client model, Windows applications (including browsers) execute on a shared Host PC. This reduces cost and improves control and manageability. As in the NC model, the zero client stations are lowest cost, secure and environmentally efficient. As in the personal computer model, the display adapter resides in the same computer as the application. This preserves performance because it eliminates the need for a network transmission protocol that degrades CPU processing and injects delays due to network overhead. When all the strengths and weaknesses of each desktop configuration alternative are considered, the zero client technology offers flexible and valuable options to users seeking minimized costs of ownership and improved control.
Article submitted and promoted by : http://www.sundenc.com