Slashdot Log In
New York Times Says Thin Clients Are Making a Comeback
Posted by
timothy
on Mon Oct 13, 2008 01:01 AM
from the dialectic-materialism dept.
from the dialectic-materialism dept.
One of the seemingly eternal questions in managing personal computers within organizations is whether to centralize computing power (making it easy to upgrade or secure The One True Computer, and its data), or push the power out toward the edges, where an individual user isn't crippled because a server at the other side of the network is down, or if the network itself is unreliable. Despite the ever-increasing power of personal computers, the New York Times reports that the concept of making individual users' screens portals (smart ones) to bigger iron elsewhere on the network is making a comeback.
Related Stories
[+]
Office-Worker Linux: It's Here and It Works 615 comments
A few weeks ago, dot.kde.org featured a great why-should-this-be-amazing story about Linux being used as the day-to-day desktop operating system for city employees in Largo, Florida. Roblimo got a chance to see the system in action to find out how ordinary office workers are proving that the old "Linux is tough to use" shibboleth is nothing but FUD, and how a medium-sized city is saving buckets of money by minimizing the tax dollars spent on licenses and hardware. Oh, and they've also pre-empted the kind of costs (in hassle and money) that can face any organization that Microsoft suspects may have some licenses out of order. This is the kind of thing every elected official should have politely waved in his or her face by concerned taxpayers. The Largo system uses KDE on Red Hat, but since both KDE and Gnome are paying much attention to user interface, similar systems could easily be running on various combinations of hardware / distribution / desktop system.
This discussion has been archived.
No new comments can be posted.
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
Full
Abbreviated
Hidden
Loading... please wait.
How cool! (Score:5, Funny)
Now, the terminals that work has had since 2003 are back in vogue. Awesome.
Re: (Score:3, Insightful)
Could have told you that was coming (Score:5, Insightful)
Re:Could have told you that was coming (Score:4, Interesting)
Parent
Re:Could have told you that was coming (Score:5, Informative)
Which is why its not a great idea putting mission critical thin clients across a WAN
Though having worked for several years in large corporate environments (and their associated love for citrix farms), I would observe
- WAN accelerators work. A riverbed (mind you at ~$50,000AUD a pop it ain't exactly cheap) will make a 2M link seem like LAN speeds for the protocols its optimised for. Depending on cost of bandwidth....
- Consolidation does not have to go overboard. If there are at least a few hundred users, it can be cost efficient to run a local server. Most network problems that are not a result of a bungled change / cabling stuffup are WAN.
- Government network? good luck with that buddy!
- The bean counters find it very easy to quantify the cost 'savings' and push their agenda as such. However for your potential losses due to downtime caused by network outages.... heck the fortune 500 I am contracted to presently doesn't even have a method for estimating the dollar cost of downtime, let alone a method for estimating the amount of downtime likely to occur (needless to say they also choose the cheapest carrier, which has a ridiculous inability to meet SLA, and then consolidate like mad to place even more reliance on the WAN).
Like most things in IT there is no silver bullet or magic formula, each case needs to be judged on their own merit.
And on a side note, given how much hardware costs have dropped and the fact that user requirements have remained relatively static (i.e. most generic office workers are still using the same software as 4 years ago), how hard can it be to embed the email client (with local cache so they can at least view emails they already downloaded) and office suite on the thin client itself so at least they can keep working on documents?
Parent
Re:Could have told you that was coming (Score:4, Interesting)
here is the kicker you can't easily run citirix and windows apps across a WAN. too much bandwidth that is lag sensitive.
my company runs an AIX server with ssh access. each user literally SSH's into the server which loads up the acccess to the point of sale/inventory database. Everything important is tightly controlled. but the fact that you can run it over a dial up 36.6 modem effectively means that even if the internet is choking you can still work.
Parent
Re: (Score:2, Informative)
Re: (Score:2)
Re:Could have told you that was coming (Score:4, Insightful)
At this point, if the network goes down then all clients, thin or thick, will effectively stop working anyway.
Parent
Re: (Score:3, Insightful)
Re: (Score:2, Informative)
Re: (Score:2)
You sidestepped the major issues and questions resolving thin clients and related setups:
The first question is: for your supposed all-around solution, what exactly is it intended to be used for?
The second is:
Why could said solution in the first not be solved by people having computers in the first place, albeit cheap ones if stuff is so minimal it can be done on thin client?
Re:Could have told you that was coming (Score:4, Interesting)
As an example, SunRays [zdnet.com] generally scale much better than a cheap PC environment, with much better return on investment.
You are going to be spending money on servers either way. According to your own figures, you have 7.5 users per server. SunRay solutions typically yield 20+ users per server cpu core. I'm not doubting your figures, but what do you guys do that requires so much back-end power? Are they multi-cpu servers? Fully utilized? Are they under-utilized? Single or dual cpu servers? Obviously, I'm not in your position, but before I looked at desktop solutions, I'd look at server consolidation. VMware or similar might save you a bundle and make things easier to admin.
As for new software, SunRay environments are pretty easy to patch and deploy new software in. As a matter of fact, that's one of the strengths - deploy the patch or app to a single server or a few servers, and you are done.
Electricity is hardly a selling point if you're losing productivity and still spending the money on servers, to boot.
Obviously, achieving functionality is more important than being efficient. However, the point of thin clients is that they generally keep office productivity the same or better, IT efficiency is tremendous, and the equation ((thin clients * users) + (servers)) is less than ((full PC desktop) + (servers)) generally holds true. At that point, saving several hundred KwH might be pretty attractive.
Parent
Re:Could have told you that was coming (Score:5, Interesting)
Seven hundred?!?! Microsoft had a web page where you could put in your client requirements and they would tell you how many Win 2003 TS machines you would need to support these clients. I don't think we ever got it down to fewer than 10 users per server - how did you manage 700?
Currently we have four servers for about forty seats in our labs. They don't get much usage, and people don't seem to notice they're sharing a machine with the other 10 people on that row of the lab.
I'd give thin clients to everyone, but then someone in an office of their own will tell us they really need Skype, and they really need a web camera... I suppose these things could be connected to a thin client and forwarded over USB, but it's not something we've tried...
The other show-stopper is where users need admin rights for particular software. It does still seem to happen, mostly with big important pieces of software like our finance system or student records management. It may just be it needs to write to the C: drive so we could bodge it with access rights, but we don't want to screw up the installation so the user gets admin rights. Now, could we do that on a shared Windows 2003 TS box? I don't think so. With VM tech we could give them a VM of their own to play with though...
VM tech has also helped us deploy Linux and Windows to our labs. Previously we had say four servers running Linux and four running Windows, and if the lab session needed Windows then there were four Linux servers sitting idle, and the users crammed onto the four Windows servers. With VMs, we stick a Windows and a Linux VM on each server, then the users are more spread onto the eight servers. Win.
Parent
Eh (Score:5, Insightful)
There are plenty of downsides too. While it might be easier to maintain, it is also easier to fuck up. Someone does something that breaks a piece of software, now the whole department/company/whatever doesn't have it rather than just that person. A network outage is now a complete work stopping event rather than an inconvenience. Special software installs for special tasks are hard since that software has to be tested to make sure it doesn't hose the server.
I could keep going, if I wished. Now that isn't to say that means the thin client model is bad. In fact we are hoping to do it for our instructional labs at some point. What I'd really like (and there are VM solutions to do) is that not only would we have thin clients, but a student could use a laptop as a thin client too and load our image from their home or whatever.
However, the idea that they are just cheaper/better is a false one. They can be cheaper in some cases, in others you can easily spend more. Likewise they can simplify some thing and make others more complex.
There isn't a "right" answer between large central infrastructure and small distributed infrastructure. It really depends on the situation.
All I will say is if you are looking at doing this at your work as you suggest be very, very careful. Make sure you've really done your homework on it, and make sure you've done extensive testing. I don't think it's a bad idea, but be sure you know what you are getting in to. Just remember that while people get whiny when, say, an e-mail server goes down, if the terminal server goes down and NOTHING works, well then people go from whiny to furious in a second.
It's the same kind of deal with virtualization. It is wonderful being able to stack a bunch of logical servers on to one physical server. However if that one physical server dies you can be way more fucked. You have to spend a good deal more time and money in making sure there is proper redundancy and backups and such. So while packing 10 servers on 1 using VMWare Server (free) might be nice and cheap, you also might be creating a ticking time bomb. You then might discover that putting those 10 servers on a small cluster with a fibre channel disk array and VMWare Virtual Infrastructure (not free) solves the reliability problem nicely, but isn't quite as cheap as you thought.
Just something to be careful with. At work we have both sorts of things. We've got individual desktops, and we've got thin clients (though we actually got rid of most of those). We've got individual servers, we've got virtual servers, and so on. All methods have advantages and disadvantages. I am not a zealot either way, just warning that a change from a decentralized to a heavily centralized infrastructure isn't something to be done lightly. You solve various problems, but introduce a host of new ones.
In particular hardware reliability is something you want to keep in mind. You for sure want an "N+1" situation with your terminal servers, perhaps even more than that. You can't count on the hardware being reliable. Hopefully it is, but I've seen even the real expensive, redundant shit (like a Sun v880) fail with no warning. When it's the be all, end all and all work stops when it is down, that just can't happen.
Parent
Actually, it's probably a PR story (Score:5, Insightful)
1. Actually, regardless of whether they are making a comeback or not, or what their advantages and disadvantages may be, this is probably just a PR story. Just like the "The Suit Is Back!" that got traced back to a PR agency a couple of years ago.
PR loves to masquerade as news because it bypasses your BS filter. An ad for Men's Warehouse suits gets looked over, a piece of news that you won't get hired unless you wear a suit, tries to replace your premises with theirs and let you take a leap to the "I must buy a suit" conclusion. Or better yet, to the even better conclusion, "I must only hire people in suits 'cause everyone else is doing it." There are a lot of sheeple out there who only need a "The Herd Is That Way -->" sign to willingly enter someone's pen and be sheared like "everyone else".
For anyone who's not sheeple, this is a non-story. Whether _you_ need a server instead of PCs or not, depends on what _your_ needs are and what _your_ employees are doing. Use your own head.
The only ones who need an "everyone else is doing X" story are those who have to follow a herd to feel secure.
Hence, the love PR has for this kind of story.
2. Over-simplifications like "all they need is internet, database access, and word processing" were false when arguing why grandma should only need an old 486, and tend to be just as false for a company. So you'll have to do some analysis if for a particular company that is indeed true, or just glossing over what's really going on. (Or even wishful thinking by some IT guy who feels his job would sound more important if he was overseeing a server.)
E.g., a lot of companies have salesmen who go with a laptop to various customers to give a presentation and try to win a contract. Are you ready for the case when that guy you're trying to sell insurance doesn't have internet to connect to your server via VPN? Are you sure that those server side apps' files can be converted flawlessly to MS Office or whatever those sales guys have on their laptop?
It's just one example where goimng, "bah, they only use database apps and word processing" is glossing over a more complex problem.
3. The argument for saving costs is a good one, and far from me to advise wasting money. But you have to be sure that you're actually _saving_ money across the organisation, not just saving $1000 in the narrow slice you see, at the cost of causing $1,000,000 to be lost in workarounds and lost productivity somewhere else. Entirely too much "cost cutting" lately is the latter kind of bullshit theatre.
E.g., if someone costs you $100,000 per year -- and I don't mean just wage, but also electricity costs, building rent, etc -- saving $1000 is nullified if it drops their productivity by as little as 1%. Saving a few hours per year of an IT guy's work can be a very bad trade off, if it costs that guy as little as 5 minutes total per 8h work day to put up with the quirks and delays of the centralized system. (480 minutes a day, times 1% is 4.8 minutes.) It can add up very easily to that. It only takes wasting 1 second per form through some web-app instead of letting that guy massage the data locally in Excel or Access(*), to add up to more than that in a day. A close enough approximation can very easily be approximative enough to actually turn the whole thing into a loss.
(*) ... or whatever F/OSS equivalents you prefer. This is not MS advocacy, so fill in the blanks with whatever you prefer.
And as you move higher up the totem pole, things get even funkier. If a salesman is doing contracts worth millions of dollars with those presentation, I hope you better save a _lot_ with that centralized solution, because it only takes one lost contract (e.g., because he couldn't connect) to put a big minus in the equation. E.g., if you're going to pay a CEO tens of millions per year, and actually believe that his work is worth every cent (heh, I know, but let's keep pretending,) then... again, you better be damned sure that you don't drop _his_
Parent
Re: (Score:3, Interesting)
Yes,because all it takes is walking into one of those horror story networks to make you appreciate a sane network. I went with a couple of buds to help out a school chum fix a mangled network. This is in a very expensive LAW OFFICE mind you,not some "Rick's used cars" kind of place. When we get there he takes us to the server room he says " I had to show you guys this first because.....DAMN!"
The guy that came before had three HOME MADE GAMER RIGS running as servers,and of course not even the motherboards we
Re: (Score:3, Insightful)
You should be asking "Why?" (Score:5, Interesting)
Easy, same way I handle it at our office with our terminal server: "You can't do that."
Employees have no business copying CDs worth of data to (or worse, from) the office. In the eight years since the implementation of our terminal server environment, I have had exactly zero cases where there was a legitimate need to copy large amounts of data from the terminal server.
Your computer at work is for working, not playing games when you think nobody is watching. Almost all of the complaints I get from employees wanting a "real PC" instead of a thin client revolve around their desire to screw around on the clock without being detected.
In 100% of the cases where the employee was granted a PC instead of a terminal, later investigation revealed unauthorized usage within one month, ranging from forging call sheets to play flash games to a salesman using over 75% of the company's total internet transfer in one month at myspace.
Parent
First Post! (Score:5, Funny)
...or, well, it would have been first if I wasn't on a thin client waiting 15 ^%*^&# seconds for a keystroke echo.
Thin clients? (Score:2)
They're still waiting for the cancel/allow box to show up
Another cycle in the industry (Score:5, Insightful)
Re: (Score:2)
Yay! People rediscover the advantages of thin clients! How long until they rediscover the downsides...
That would be when the vendors have made their cut and hand over to the consultants for their turn at the trough.
Re: (Score:2)
This isn't really an industry cycle, it looks more like a plug for a bunch of current products, ala: http://www.paulgraham.com/submarine.html [paulgraham.com]
Re: (Score:3, Insightful)
it's not about rediscovering the advantages/disadvantages of thin clients. AFAIK thin clients were never fully abandoned. it's simply about finding the right niche for thin clients.
for instance, if you're setting up some computers at a public library that only need to search through the library catalog and nothing else, then thin clients are the clear way to go. if you're running a school network where thousands of students will be sharing a few hundred computers, but they'll need word processing, desktop p
Re:Another cycle in the industry (Score:4, Interesting)
the computing demands of the casual user hasn't increased that much since the days of Windows 95
Right, just try watching YouTube on Firefox with a Pentium 133.
by giving everyone else thin clients, you'll give them less chance to screw up their system, thus giving them more uptime and more reliability, which users will appreciate.
Uh huh, you can solve the "chance to screw up their system" by keeping the thick client but virtualising the OS, and as for more uptime and reliability it will only be as reliable and uptimely as is your network/servers, which is in most contexts probably not any better, plus you have to deal with general downtimes, and this way people are going to end up with all their eggs in the same basket, which, although avoidably, could bring huge IT catastrophes. Relying entirely on a centralised network is absolute madness, a single network administrator's mistake, a lack of redundancy combined with a hardware failure, a bad decision or incompetence could paralyse an entire infrastructure. Centralising everything only looks nice on paper.
Parent
Re: (Score:3, Insightful)
>Even with a GUI terminal, if it was stripped down and wasn't Windows based
>(and had drastically limited Internet access), I think a lot more would get done around offices.
Bingo! That is exactly what we have- Linux server, Linux apps, Linux thin clients (160). Everything is locked down tight. We have everything users need in order to be productive and nothing else (accounting apps, OpenOffice, Firefox, Sylpheed, IceWM, some utils). Internet access is only through a white list of approved sites. B
That "true computer"... (Score:5, Interesting)
It's annoying as hell, but if my network craps itself I still have a working computer in front of me and I can still do a subset of what I was doing before. Not so with thin clients.
<tinfoil mode>
Of course they want to take the actual computer away from you, they want to have control over you. If they could, your "computer" would be a mindless terminal to a Big Brother Approved mainframe that spied on everything you did.
</tinfoil mode>
Re:That "true computer"... (Score:5, Interesting)
The Linux Terminal Server Project [ltsp.org] is actually pretty good. And useful for a variety of things beyond just saving dough on the desktop end. Remote access is one that comes to mind. Sure, you could have a bunch of X terms, but this will work with ANY box with a PXE (hell even Netboot) NIC. You don't need virtualization or any of that garbage. UNIX was designed as a "multi-luser" operating system ;), back when mainframes were last in vogue. Xwindows is really quite good over a slow network and has been for DECADES.
Now, I want to stress that I am a proponent of terminals in only certain areas. A public library computer bank. A factory environment, where you want your server safe and securely away from sparks and heat. A customer service environment where the employee is only doing one or two things. My business ops people would have real computers for the reasons you mentioned. I want them to be accounting and developing even if the server is down.
Parent
Re: (Score:2)
Correction, you can also boot using Floppy, CD, or USB boot image.
Thin clients ... (Score:2)
Dumb clients, fat clients, thin servers, retarded paywalls.
It's simple business: (Score:5, Insightful)
When you have flogged off all of your customers with a thin client, the new thing is a "better-er-er" thick client.
Whole thing sounds like very simple 101 style marketing. Why try to sell someone something they have? Convince them what you have is better. Total no-brainer imo.
Necessary (Score:2)
Middle ground? (Score:4, Insightful)
How about a netbook-style device which could offer limited functionality on it's own for email, web, basic office apps (say a boot image updated from the central server when connected), and used as a thin client at the office plugged into a docking station with proper display(s) and keyboard+mouse? Best of both worlds?
Re: (Score:3, Insightful)
Hmm, you mean, like one of those laptop things? /snicker
Re:Middle ground? (Score:4, Funny)
note: you have to turn aero on for a complete thin client experience
Parent
Re:Middle ground? (Score:4, Insightful)
How about a netbook-style device which could offer limited functionality on it's own for email, web, basic office apps (say a boot image updated from the central server when connected), and used as a thin client at the office plugged into a docking station with proper display(s) and keyboard+mouse? Best of both worlds?
Why, all you'd need is some kind of Window System that could display X, where X could be any number of applications.
Parent
Happy coincidence, Thin Client & Virtualizatio (Score:2, Interesting)
We have recently adopted a phased approach of deploying new thin clients as our estate of traditional desktops hit retirement. After having seen several false dawns and uncomfortably proprietary solutions in the last 15 years, it was only now that we have been happy enough with the whole solution (thin client HW, network connectivity, back-end virtualization SW) to take the plunge.
There are now a range of HW clients (we use ChipPC [chippc.com]).
There are a couple of viable virtualization systems (we use Citrix Xen [xensource.com], with
Struggling economy, phooey! (Score:2, Funny)
Finally I can sell all the Wyse 120 terminals I have in the garage! If you want me I'll be high-rolling at the casino for a couple of weeks...
My clients are fat? (Score:5, Funny)
My clients are all obese, and show no intentions of slimming down; what am I doing wrong?
Re:My clients are fat? (Score:4, Funny)
You're working in America.
Parent
2009: Thin client v Linux on the Desktop (Score:3, Funny)
Oh yes its back the battle that everyone has been waiting for its the Rumble on the Desktop, the fight of the century, the challenger is the undisputed next year champion, fighting out of California by way of Finland it is the Penguin himself, Tux "next year" Linux.
And now the champion, dominating in the 70s, losing form in the 80s, disappeared as a recluse in the 90s and the start of the century but now he is back to claim his crown. With the black trunks and green trim its Thin "Latency is a Bitch" Client.
Lets have a good clean fight to finally decide who will be declared the Desktop champion of 2009.
This fight is sanctioned by the ODC (Optimistic Desktop Council) and will be fought under rules of low data, huge assumptions and a complete lack of understanding on the total size of the market.
Re: (Score:3, Interesting)
This isn't boxing, more like wrestling. So don't be surprised if you see VM "you trashed your OS here have this backup virtual image" ware jump up on the ring, headbutt in all directions and virtualise the shit out of your thick clients.
Am I the only one who believes that the future is not in thin clients but in desktop supervisors who make all your OSes run transparently virtualised? I'm talking about 10-15 years.
I support the return of thin client entirely (Score:3, Insightful)
too many dumb users (ok I am being too harsh here, too many uneducated users) these days. Thin clients = less freedom, which in case of most users means they'll make fewer mess ups.
This means less boring maintenance work for IT people, in large companies especially.
Flash Kills Thin Clients (Score:2)
Unless you've got a lot of bandwidth to spare, Flash will kill performance.
Yes, i observed that (Score:2)
there was a short period bridging the vt100 terminals to the sunrays from 1997 to 2000, where the University library installed personal computers for accessing their network.
No, seriously. This is non-news.
The transition to personal Computers stopped long ago. I can not remember to have seen an institution in the last five years switching to a PC-based infrastructure, but i see since approx. 2001 a rise of thin clients in larde organizations. The organizations for which this pays off will get smaller and sm
But... (Score:2)
New York Times Says Thin Clients Are Making a Comeback
But in Texas they're as fat as ever.
About 7 years ago I heard the same story (Score:3, Funny)
After announcing the loss and accompanying layoffs, he actually followed it by saying "And I don't think suggesting thin clients will help us out of this one."
Man, it was so hard to keep from laughing...next time I hear that, and it sounds like I will hear that again, I think I'll just risk my job and have a big belly laugh.
it follows from physics (Score:4, Insightful)
From physics, it's obvious that centralized computing is more energy efficient than distributed one. The longer distance you have to move energy (that encodes the information) to compute the results, the more energy you need. Also, centralization allows for better resource sharing.
The only issue is who pays for the costs. Mass production of computers allowed to decrease their costs to the point that distributed systems were cheaper than centralized ones. However, as the demand for computer power grows, energy spent on computing itself enters the equation, and the times will change again.
Too bad they suck (Score:3, Insightful)
I worked exclusively through thin clients for a year at my last job and absolutely hated it.
It was slow, and ungainly and every now and then - from a few hours to a couple of months - someone else's X session windows would pop up on my screen. Wonderful in an environment where we worked with secret (as in classified as) information. We knew the problem, and the IT guys could usually fix it in a few minutes, but the fix always seemed to be temporary somehow.
Not to mention you're costing productivity for people like me who tend to work very rapidly via esoteric hotkeys, and rapid fire keystrokes, and using the keyboard buffer to issue commands to dialogs, context menus, windows that haven't yet appeared. One of my earliest employers once described seeing me work at a computer as "really making that thing sing". So sticking me on a slow machine or dumb terminal is costing you my productivity and happiness. And it's not like a decent machine $1500-2000 is really that big of a deal spread out over the several years it will last. Especially if it's one more straw kept off of the camel's back that keeps me for looking for another job and costing you domain knowledge and experience with your unique problems when I leave.
IMO, thin clients should be reserved for "guest" users who will only be temporarily using your network where no degree of customization or where speed is not important. Like an interactive presentation or a library, or some temporary event.