The Nivo is a new ‘ultra thin’ client device under development by British, not-for-profit organization Ndiyo. Its principle intended market is developing nations and the project is based upon a non-profit model. The aim of the project is to create hardware and software that can lower both financial and technical skill cost of computer service provision.
Other companies have experimented with the thin client approach but the Ndiyo takes things a step further in terms of miniaturisation and streamlined, single purpose design, hence the use of the term ‘ultra thin’.
At the current stage, the project consists of Linux based software and prototype hardware. In this article I’m going to explore some of the issues surrounding thin clients in order to paint a picture of where a device such as the Nivo sits.
The thin client concept
For those of unfamiliar with what a thin client actually is, the basic idea is this:
In order to operate, a typical desktop PC requires only a source of power. It can be connected to network such as the Internet or an office network but it does not require a network of other computers in order to work. The independent operation of a single PC is made possible by the fact that it has a complete set of component parts needed to support its basic functions; these components include a logical and computation facility (the CPU), short term storage (the RAM), and long term storage (the hard disk).
The thin client takes a different approach. In a thin client, as many of the software and hardware components as possible are moved out of the client computer and onto a server. Simplifying the client machine in this way reduces its size, power consumption, cost, and maintenance requirements.
So, to shift the balance of components from desktop machine to the server is to move closer to the thin client ideal.
You’ve probably had a taste of the ‘thinning out’ of a client system if you have ever worked in a large office: Often, in such an environment, it is considered inefficient to give each computer its own printer. In such cases, a single, shared printer may be located in a common area of the office. To consider the idea of thin client adoption is to pose a question as to whether this idea can be taken a step further: How much can you remove before the computer cannot be used for typical office tasks? In most cases, removing the display and the input device probably exceeds a sensible upper limit of how far you can go along the thin client route. Computers without input methods or without a display do exist and are used but they cannot replace the functionality of a desktop PC to any meaningful degree.
Thin client adoption
Some of the computer industry heavyweights such as Sun and Oracle have experimented with the approach in the past but without great success. As a result, in most people’s mind, the thin client concept will always be associated with a series of early 90’s network based computers that never really took off.
In fact, the story of the thin client goes back far further than this. The earliest thin clients were the dumb terminals used to access early mainframe computers. Such early computers were both enormous and hugely expensive. In order to extract the greatest utility from each mainframe, dumb terminals – computers with limited processing power and no local storage – were created so that a group of people could all share the resources of a single mainframe computer.
The thin client existed before the desktop PC; the desktop PC killed off the thin client.
I would argue that some of the difficulties that have frustrated previous creators of thin client solutions have been of a psychological rather than purely technical nature. The desktop PC is the dominant model and consumers need a considerable shove before they are willing to shift over to a new paradigm. It’s possible to draw a parallel between thin-client adoption and public transportation adoption. In the case of the latter, there are cases where switching to public transport is a sensible idea, but the bottom line is that it would take a lot to get some people out of their car. Boundaries, of pride of ownership, individualistic appeal of independence, and feelings of security, have to be crossed before people will give mind-share to something that opposes the entrenched model of usage.
Of course, there are developing nations that don’t have much of an established IT tradition, and in these environments, thin client evangelism might actually face less resistance. In a case in which IT provision is being introduced to an an institution for the first time, there are no desktop PCs to pry from the clutches from crying, pleading office workers.
Advantages of thin clients
The main advantages of a thin client solution are:
- reduced cost per unit
- reduced power consumption
- improved reliability due to simplified hardware
- single point of maintenance (the server)
- greater admin control
The single point of maintenance of a thin client is an advantage that is easily overlooked. Typically, in the case of an office full of standard PCs, the skills cost of each workstation might be a considerable fraction of the total cost of each workstation. In the case of a technical failure, time spent maintaining a single machine can range from a few minutes right up to an hour or more. Evaluating the monetary cost of a single workstation while ignoring the admin-time cost of a solution is an easy, and very common, mistake.
The main shortcomings of thin clients are:
- Single point of vulnerability (if the server goes down, all
of the
clients go down) - Performance: The maximum capability of each client machine
can never
exceed the total capability of the server divided by the demands being
made by the other clients. So, if five clients are making maximum
demands upon the server at the same moment, the capability of each
client is 1/5 of the total capacity of that server. - Capability: Some applications are not viable over a thin
client network.
For example, if I were tasked with specifying the hardware for a video
editing suite, my thoughts wouldn’t, given the current state of the
technology, turn immediately to thin client solutions.
The Nivo: an ultra-thin client
Those who have read my other articles will know that I am a sucker for ideas that challenge the conventions of desktop computer use, and so I was intrigued when I heard about Ndiyo, a British organization who are developing a series of thin clients. In fact, they refer to their system as ‘ultra-thin’ because they take the streamlining concept further than most previous attempts by other companies.
The Ndiyo Nivo (network in, video out) client contains no local storage capacity, minimal connectivity and minimal on board processing power. The unit itself takes the form of a sturdy looking, metal box with connector ports distributed on either end. VGA on one end and keyboard, mouse, Ethernet, and power on the other end. Internally, there are no moving parts.
Because the Nivo is truly a thin client, it cannot be operated in isolation; a server is required. The server is a desktop PC with the Nivo software loaded. The software works under Linux or Windows. Ndiyo even provide a bootable CDROM image to make possible a zero install setup. This image is based on the latest Ubuntu.
Software-wise, each user is given access a set of default application programs that one would expect from an Ubuntu-powered desktop such as: web browsing (Firefox), office productivity (Open Office), Email and PIM (Evolution). Remember, you’re not restricted to any particular software; you can add anything that you could add to any other Linux system. Obviously there will be limitations in terms of what the system can manage to run over a network; it wouldn’t be reasonable to expect a system such as this to be able to handle a graphically intensive, high-end game, for example.
So, to recap, once the server is up and running, the setup procedure for a new Nivo workstation would be simply to connect keyboard, mouse, and monitor to the Nivo and then connect the Nivo to the Ethernet network.
Stateless operation
Each individual Nivo is stateless in operation; that is, it contains no local storage. This means that you could, in theory, disconnect a machine while it is in use, move it into another room, reconnect to the network, and recommence work at exactly the point at which you left things: all of the applications and windows would be in exactly the state in which you left them because their state exists on the server. This feature is useful in countries with a poor energy infrastructure because it means that if the server is connected to a UPS, a temporary power cut need not be a disaster for the client users.
Future options
A home version?
The provision of basic web and office application services to a network of client machines in developing nations is a logical application of the thin client concept. However, I wonder if Ndiyo are failing to exploit the potential of their device in the consumer market of first world countries? I doubt that I’m the only person who would be interested in a home version of the Nivo; as soon as I started poking around the website I started thinking of things that I could use a Nivo for around the house.
Let’s dream a little…
Such a home version could feature, in addition to standard layout, composite video out and sound output. USB might better be a better fit than PS/2 keyboard and mouse. Such a device would make an ideal living room or kitchen computer. There are lots of such environments where one might prefer a terminal that gives access to some standard tools in a robust, low-maintenance package instead of a full desktop or laptop PC. It would be great for light use such as looking things up on the web, a bit of IRC chat while cooking, and checking the progress of downloads, etc. Using it as a music and video jukebox is also within the realms of possibility.
Regardless of how ‘easy’ it is to, for example, setup a computer in the guest room when people are coming over to stay, it’s still seems like hassle compared the simplicity that a Nivo could offer. In addition, I could imagine that many a geek might enjoy a few points of access, dotted around his or her geek lair, and all with a single point of maintenance.
The office version?
I wonder if the offices of first world nations could benefit from what the Nivo could offer? The truth is that, for 90% of ‘typical office use’, a complete PC is overkill. Most office workers use only a small number of very standard office applications. Theoretically, a ‘perfect thin client’ is just what offices need.
Having just been asked to, temporarily, add an extra Internet ready PC to the conference room, I bet many an admin has wished that he had access to something a bit like the Nivo.
Conclusion
Within the business world, the thin client faces many of the same adoption problems that are common to other non standard solutions, such as open source software alternatives, that are battling to break through into the mainstream. The famous IT maxim that ‘No one ever got fired for buying an IBM’ could be expanded and recast as ‘No one ever has the guts to bet the IT part of their business on something new and untried.’
Find out more at about the Nivo and other projects by visiting the Ndiyo website.
About the author:
Michael Reed is so geeky that he can use phrases such as ‘kitchen computer’ and ‘amusing assembly language anecdote’ without even a hint of irony. Read more about his geekyness on his website.
I see it as the way of better utilising the power we get with nowadays computers. It is so unefficient to use monsters with gigs of RAM and gigaherzes of computing power to just write a letter, listen to music and so on. It could be used in much better ways. I can imagine offices or classrooms with these filled with these thinies…
I completely agree. There is a lot of wasted power out there. I’m currently using a Toshiba Satellite 300CDS notebook computer (which is roughly 10 years old) as an intranet web-server and web-controlled jukebox. It has been working great for months now. It uses roughly 10W of power with the LCD off and the battery fully charged, which is acceptable by my standards.
If apple had an affordable thin-client system, I’d spring for an xserve or macpro, put it in a rack in my coat / wire closet and be all done with PC’s on desks.
Unfortunately, the linux offerings don’t have the software that I and my family would require in order to do that. Or else it’d already be done.
Apple does have a relatively affordable thin client. It’s $299, a bit steep for a thin client, but it has NVIDIA graphics, HDMI with HD audio, 40GB hard disk, and 802.11n, which are high-end features for a thin client.
The problem is that they don’t want you to use it as a thin client. They call it AppleTV. It’s for playing iTunes content on your TV. But it could have been a thin client that extends your Mac onto any display in your home.
But thankfully, those monster machines with gigs of RAM and GHz of computing power that we use to write software, encode video, host email/web/db servers, and frag our buddies scattered hither and yon across the intraweb can also be used for streaming music and writing emails and letters.
Now a days, even geeks, can better consolidate their machines. That’s what I do on my Mac Pro. Its hosts EVERYTHING now.
Development databases, app server(s), email server, web server(s), music and video library.
4 cores of love, 4GB of RAM, revolting amount of disk space.
Some servers run native on the machine, others run in Linux or Solaris under Parallels. The VM tech is the hot tip here.
Got a “spare heater^WPC” running linux and email server? Migrate it. Create a VM, and copy your install over. Give it a 128M or so.
Consolidate these “little servers”, get rid of the wall wart forest, or the cheap power strips, or the aged Pentiums in the the half open cases.
Yup, it’s a single point of failure, but so is the power in my house. Most of these services are lifestyle services, and not life or death services. I can handle my email being down a day.
Consolidation to a single server does not scare me.
And that’s what any thin client project is doing as well (punting on the potential “thin client talking to cluster” meme).
Now add in that new ASUS flash based laptop with wireless for a cheap, wireless, low power, thin client display.
Now if I can only figure out how to get a thin client to talk to my main Mac. I’d love to be able to log and use my Mac from an arbitrary client. Not clear if that’s what Remote Desktop is for or not. I can always X Windows in to one of the Solaris VMs, but..it’s not the same.
You are clearly not an average user, or classroom child at which this adequate technology is aimed at.
I dont think that sentence is correct:
“Performance: The maximum capability of each client machine can never exceed the total capability of the server divided by the demands being made by the other clients. So, if five clients are making maximum demands upon the server at the same moment, the capability of each client is 1/5 of the total capacity of that server.”
I dont know exactly the software solution of Nivo, but my experience with LTSP is different. Even with 10 users logged on a server, it is possible to use near to 100% of the CPU. It is normal to use a computer and not use much CPU cicles. For example, how much CPU am I using right now, writing that text? Near zero. Most of our everyday work uses near zero CPU power. So, it is possible to use near the total power of a server (meaning CPU and RAM capacity).
Personally I think the articles explanation made more sense than yours did.
I’m really not quite sure what your point was let alone how that differs from the articles.
This is a neat idea but I would field test it before proclaiming it to the world. Field test the idea that a thin client could provide break through services to 3rd world and developing nations, that is.
although having smaller hardware like that would be nice, everything the author talks about is already possible with a linux box using nfs and dhcp. its how my office is setup. and the hardware was practically free. just used a bunch of old systems and got new monitors for them. they boot and run rdesktop to login to a locked win2k3 session, where the user can only run 1 app. i have thought about doing the same thing at home, but only building a “media network” with mythtv and a bunch of diskless systems.
Your hardware may be free, but the power it uses is not. Perhaps if this device is everything the developer and the reviewer claim it is, you could begin to replace your old systems as they die off with these boxes. Depending on how much the hardware will sell for, they would probably pay for themselves very soon with the energy saved. Noise and heat would become less of an issue as well.
noise and heat are not an issue. its a doctor’s office and the systems are in exam rooms, but they are locked in cabinets so they can’t be accessed by the patients. as for power, i don’t know. i disconnected the hds and cd drives. all the systems are p3s and only have a video card, ram, ethernet card, and floppy drive to boot from.
The power savings would probably still be impressive; the system you describe most likely draws 200 watts or so just sitting idle. I guess it all depends on whether the money saved in power bills would ever make up for the cost of replacing otherwise free hardware.
You are overestimating by far!
I have measured my computer with 160 W peak and 140W idle.
The machine contains:
1.4GHz AMD Athlon
500 MB RAM
2 Hard drives, 30 + 80 GB
Network card
Old graphic card (without any cooling device)
TV tuner card
sound card
so if he uses some old P3 with 300 MHz without hard drives and graphic card I would be astounded if it sucked up more than 60W.
Is this basically a hardware VNC in a box? Looks like its a project run by Quentin Stafford-Fraser, trojan coffee pot fame
http://www.cl.cam.ac.uk/coffee/qsf/coffee.html
co-inventor of VNC
http://en.wikipedia.org/wiki/Virtual_Network_Computing
and recently a founder of displayLink
http://www.displayLink.com.
There was an interesting article about Ndyio in the Guardian newspaper a couple of weeks ago, which suggests that the Nivo is more of a tiny VNC client than a thin client. I thought killermike’s article was great, but also felt it didn’t really highlight this as a major difference between the Nivo and traditional thin clients. This really is the thinest of thin.
The Guardian article also points out that they’ve had “fullscreen DVDs playing down an Ndiyo link”, which sounds very impressive to me.
I also think it’s intersting that they’re integrating this technology directly into screens. I could imagine a time when you can’t buy a monitor that doesn’t have this built in, and I think that would be great.
Anyway, here’s that article for anyone interested:
http://www.guardian.co.uk/technology/2007/aug/02/guardianweeklytech…
Doesn’t this move each clients Xserver work to the server plus needs compression ?
A LTSP server only has to use the network transparent X layer.
Does a Ndiyo server need hw to compress each stream ? or just serve fewer clients than running as a LTSP server ?.
Edited 2007-08-23 00:35
http://home.howstuffworks.com/question272.htm
If you really want to save power, adjust your thermostat.
Well, very nice article, but I had to laugh at some of the points.
First of all: Heat pump. If you really have a heat pump, then you have it to save energy. If you put one kWh (kiloWatthour) of electrical energy into a heat pump, then you should get at least 3 or 4 kWh of thermal energy out of it (the rest coming from the 8 degC cold earth or water you cool down to 4 degC).
So if you have a heat pump running in winter with 15 kW of electrical power consumption, this would mean close to 60 kW of thermal energy (!). This would be enough to heat a large house containing approximately 10 – 15 3-room flats in the middle of a harsh winter!!! At least if this house is properly insulated (I am using the current standard for newly buildt houses in Austria, Europe).
So if you want to save energy, INSULATE YOUR HOUSE
.
The next was the refigerator. I know, that people from the USA tend to have twice as large fridges as we in Austria, but my refrigerator consumes 160W when it is running, and it is running approximately 1/4 of the time. So this would result in an average power consumption of 40 W. The deep-temperature (-18 degC) cooler is half that size and consumes probably the same, so all together they use 80 W. Even if USA fridges are twice as large, they are not supposed to have a higher average power than maybe 200 W. If they consume more (like the 1000 W mentioned), they are not properly insulated, best get a new energy-saving fridge, look at the wall thickness, then you know which one is the best insulated.
The other things are not so bad, so when exchanging every 60 W light bulb with a 20 W low-energy bulb, you can save money (because the low energy bulb lasts 5-8 times longer), you can save more money (because it consumes less energy), and you will have a brighter room (because a 20 W low-energy bulb is as bright as a 100 W conventional one). So in effect, anyone who does NOT use low-energy bulbs is just plain stupid
(at least after having read this comment).
On energy saving in my case: My household consumes very close to 3000 kWh/year, approximately 1450 of which are for electrically heating my water. So for me, the first thing to do will be to get some solar panels onto the roof.
http://prevedgame.ru/in.php?id=20508
Had a look at this and found the video compelling viewing. (Compelling …. as in I went looking for prices on screens – yet to find in UK).
It seems, however, that each user does not get their own audio connections which limits some aspects of use.
No sound is ok in an office where noise disturbs numbers of people. But at home or in a training context (using voice overs) it would be a problem.
X11 has this limitation too. Doing audio with USB is pretty standard so the issue is infrastructure which falls back to the wider community.
Perhaps this shound be X12?
This is great! Just what I have been waiting for. I love the Sun ultrathin client we use at work, but the software will not run on my OpenBSD boxen. Hopefully this will?!?!
(Please..)
‘No one ever has the guts to bet the IT part of their business on something new and untried.’
So on the other side of things, a company may well be better off by employing one admin more than they need, and allowing them some extra budget to buy stuff they just want to play around with, so that in the end they might be the first to reap the benefit of deploying new AND well – proven technology.
Because this thin-client stuff makes sense, although not so much for your typical MS-locked in small business office. You have to pay per-seat for MS Software as far as I know, even if you only install it once on a server.
I guess, any business switching to such a solution would be very tempted to avoid as much per-seat licensed software as possible.