Less than zero-client

I took a call last month from a client (let’s call him Pete) seeking advice on a proposal he was working on to replace his company’s existing desktop infrastructure. Pete had already determined that he would be replacing about 80% of his desktop PCs with zero clients, but could I walk his team through the desktop virtualization platform selection process? I was a little concerned to hear that Pete had locked down the desktop platform before confirming the core infrastructure, and suggested that he hold off any final decision until after we meet. Pete was reluctant to hold off, concerned that he would loose his desktop budget if he did not spend it before year end, but agreed provided I could come in the following week to work things through.

I juggled a few things around and we booked out a couple of days to work things through. A week later we had a validated proposal for the desktop and supporting server infrastructure; and to Pete’s evident pleasure he was able to tell me “see I told you I needed a zero client.”

So what then is a zero client, and why was Pete so interested in them? Unfortunately, as with so many marketing lead terms, the short answer to the first question appears to be that a zero client is anything that a vendor’s marketing team wants it to be. Furthermore, given the wide diversity of client technologies available today, attempting to define a rigid technical standard in the face of multiple competing vendors’ marketing initiatives is a largely pointless exercise. Better then to attempt to define a zero clients in terms of their operational characteristics, and to do that we need to start with a history lesson.

SAN JOSE, Calif.–(BUSINESS WIRE)–Dec. 5, 1995–BYTE Magazine’s selection of Wyse Technology’s Winterm for the “Best of Comdex” award is a significant step towards validating the high profile and much touted low cost network computer concept. Winterm, the world’s first Microsoft Windows terminal, and only $500 Windows computing offering available today, performs many traditional desktop computing functions at a fraction of the cost of today’s PCs. Wyse’s Winterm was lauded by BYTE in the Communications Hardware category for its uniqueness, technical innovation and strong potential for having a significant impact on business computing worldwide. In addition, BYTE and other leading trade publications recognized Winterm for being the only network computer-type offering available now which provides the simplest and most cost-effective method to deliver Microsoft Windows 3.1, 95 and NT software applications to the desktop.

When Wyse introduced the Winterm in late 1995 the standard operating procedure was to unpack it, plug it in, configure it, and forget about it. The only management it needed was to stick an asset tag on it and depreciate it over 5 years. However, over time as deployments grew in size, the need for a remote management capability became increasingly important. Wyse introduced the Wyse Remote Administrator 3000, which provided for remote configuration and firmware update capabilities. Initially thin client management was a low impact activity, firmware updates were few and far between, and some organizations managed to get by without touching the admin console for years. Over time however, as vendors moved to more richer operating systems such as Windows XP Embedded and as thin client software delivered new features, so software patches were released more frequently and so the need for a more regular maintenance grew, eroding the benefits that thin clients were purported to provide. At the same time, the capital cost advantages that thin clients offered were rapidly disappearing. The first Wyse Winterm was announced to the world as the “only $500 Windows computing offering available today,” a small fraction of the $2333 that was the average PC price in 1995. Yet today when an entry-level business PC may be had for $350 or less the average price of thin clients remains around $500. A PC replacement that is more expensive to buy than a PC and needs almost as much management is hardly a compelling sales pitch. Granted lower cost thin-clients are available, but the management challenge remains.

The zero client is then a direct attempt to address the both the high cost and the high management overhead of mainstream thin-clients. Bragging rights for the first use of the term zero client as related desktop virtualization goes to Pano Logic who introduced the term to describe a technical innovation that replaced the complexity of remote displace protocols, and thin clients running a full OS and software stack with a simple ‘brick on a pipe’ approach to desktop virtualization. The Pano Device has no operating system, no software, no drivers, it’s a pure hardware solution based on a Field-Programmable Gate Array (FPGA) that enables the server to see the brick as a remote extension to a desktop session running on a server. Pano Logic provides a good introduction to the Pano Device architecture here (pdf). Pano Logic weren’t the first to take this approach, NComputing delivered a functionally similar solution also based on a FPGA a year earlier, they just missed out on the Zero Client marketing opportunity. Microsoft has also got into the game, popularizing the use of zero clients in education in conjunction with its MultiPoint Server with HP, Wyse and others providing the hardware (although it must be understood that MultiPoint Server is very tightly focused on the education market and should not be directly compared with other zero client solutions).

So kudos to NComputing and Pano Logic for coming up with the idea and the name, and to Microsoft for recognizing and re-purposing a good idea when it comes along. All three solutions are less than or equal to the cost of the cheapest thin-client solutions, with the list price of a USB attached NComputing U170 being only $70.

While it would be most accurate to say then that the term zero client should really be reserved for this brick on a pipe architecture, like any good marketing term, ‘zero client’ has rapidly been redefined as a general term describing a next generation thin client with reduced management overhead.

Consider the following phrases that vendors have used to describe or define their ‘zero client’ solutions:

“… relatively simple, limited-function devices with operating systems that are not exposed to the end user.”

“Eliminates the need for regular desktop [driver] support by IT”

“These devices are designed specifically for a single environment and provide substantial benefits in speed, management and security, energy and user experience”

“Thin client with no local storage”

None of the devices are zero clients in the sense that a Pano Device is, they all use software/firmware running on either proprietary or off the shelf hardware and were all designed with the fundamental premise that their code would need to be updated from time to time to support new features or address security issues. But does this really matter? Arguably not; the reason why Pete was looking to a zero client solution was was not driven by desire for a specific hardware technology, but a need to address the management overhead of conventional thin-clients. He didn’t so much need a zero client just one that is virtually zero.

So where does Teradici fit in? Well, that a tough one, Teradici isn’t reliant on a FPGA instead it runs software, or rather firmware, running on an ARM based processor. Teradici publish a good background paper on the architecture of their system here that shows the layout of their TERA1100 Portal Processor chip. But even though it runs code on a processor rather than it being a collection of fixed gates I don’t think anyone who would quibble over this and suggest that this also wasn’t at its heart a zero-client system.

That being the case, if its not about the hardware what are the essential characteristics of a zero client that differentiate it from a conventional thin client?  I’ll offer the following as a first pass at attempting to define what a standardized set of requirements for a zero client might be.

• Plug and Play

Most importantly, a zero client must be a true plug and play appliance, capable of working out of the box without the need for any administrative intervention. If the installer has to do any more than plug it in, connect keyboard, mouse and monitor, record a serial number and stick on an asset tag it fails the test.

• Zero variation

A factory-fresh zero client must do more than just work. It must, from the moment it first boots, be functionally identical to every other similar device it shares a network with. The ability to ship a device pre-configured would not qualify a device for zero client status either, as the administrative overhead of ensuring that the correct configuration is available at time of shipment can be just a burdensome as configuring it once it arrives onsite.

• Zero Configuration

Once powered-on, a zero client must be able to join a network, identify and enumerate all compatible desktop/presentation virtualization environments and connect to them without further configuration.

• Zero impact updates

If updates are ever needed to support new functionality or address bugs, updates must be made without any impact to users and must not impact end user operations.

• Zero touch updates

If updates are ever needed to support new functionality or address bugs, updates must occur across the network without the need to engage support personnel, regardless of where the device is installed.

• Low update frequency

If updates are required they should only be needed to provide additional functionality or bug fixes. Frequent security patching either scheduled “Patch Tuesday” style or otherwise are not acceptable.

• No driver updates

It should not be necessary to install device drivers to support locally attached peripherals.

I think that’s it for now, if you have any comments or suggestions I’d love to hear them.