will we go back to centralized computing?

[supernova87a]

I post this here b/c I'd like to hear the opinions of informed people who hopefully will have a more solid grasp of the issues involved:

The first modern incarnations of computers were mainframe, centralized machines which shared time among many users because it was so expensive to make the things. Then we moved to personal computers because it was getting cheaper to make, and who after all wants to be tied to someone else's mainframe.

As networks get faster, here's my question: will we move back to a centralized system where our "personal computer" is just a terminal that doesn't actually do the computing, but sends and receives information from the central unit?

I've been thinking about this a little, because although it sounds silly right now, if networks get fast enough, there are some advantages.
1. Computers would be much less bulky and affordable for the individual user. Just a flat screen and a keyboard/mouse, and a network connection which you plug into your wall, linked to the main computer. No more fiddling with all the stupid cable connections, configuring the hardware, adding components. All the software you need is on the central server.
2. This would be attractive to software companies -- no more pirated software possible. Everyone uses a central service, which accounts for users and charges the fees. People who want to develop, write free software, whatever, it all goes on the central computer, and others can see/ use as they like.
3. Hardware is more efficiently used. No more individual computers sitting around useless. Just enough resources to meet user loads, and upgrade the hardware at one location when you need.
4. Storage is more efficiently used. Why does every person need their own copy of XYZ movie? Keep a few copies on the central disks, and allocate use as requested by users. Not everyone wants to watch at once... People will stop redundantly filling their disks with stuff that can be gotten at anytime from the server.

Now, this all depends on having fast connections, and maybe that'll eventually come true. There are also some big brother-like issues where a few companies might eventually own everyone's access.

But does this scenario sound possible?

 

[Elledan]

Centralized computing is used more and more in companies, but for the average (non-business) user, it has very little advantages. True, they won't be able to screw things up as many times as they do now, but what if their connection to the mainframe is cut off?

Also, if you would have the choice, would you want to store your information and run your programs on a mainframe you've never seen?

I think that people who can use multiple PC's in their home (read: geeks and amateur-scientists) might benefit from their own mainframe, which they can maintain, because it would cut down on costs and maintenance, but for people with just one PC I don't see it as something useful. PC's are already extremely cheap.

You were right when you talked about 'big brother'-like issues as I pointed out above. Storing all of your (personal) information on a central server is not such a good idea regarding privacy. Hmm... central servers... personal information... where did I hear that before? Ah yes, MSFT's .NET.

If you enjoy your privacy, things like .NET are things which must be avoided at any cost IMHO. That is, unless security, especially that of Windows OS's (do I hear some people laughing right now?) becomes 100% reliable.

 

[Turkey]

There's more to the central/distributed picture than economics, which IMO are much more powerful. People just like to own their own stuff. Reference CDs: CD sales are still strong, even though any music a person could want is available for a few minutes or hours time. Also reference cars: people drive their own cars, even when buses are cheaper.

Plus, it would be almost impossible to run this system without restrictions on software installations. So the scenario would be you find this great new piece of commercial software XYZ which came out yesterday and you want to try it (let's say it's Nero Version 6). Oh wait, you don't have a CD burner. Hm... let's say it's Rational Rose Version 17 (I don't know what version it's at right now). Well who's heard of Rational Rose? Next to no one, which is why it's not installed on the centralized computer. But it's a $3000 program. Who pays for it? Maybe you have to have v17 and are willing to pay $3k for it. What about the next guy who wants to use v17? $3k also? I don't think so. What if everyone pays per use? Then v17 has to cost more than v16, unless you want to force upgrades. Plus, I want to use v17 today. But since it's not installed, I have to request it. "That program hasn't been tested and validated. It'll be two months before it's available." Or worse "That program is incompatible with our system. You'll have to find another." Actually no, you'll have find a new customer. The alternate scenario is that the service has partnerships with certain stores, so if you want a program that isn't installed, then you go to these partner stores (at which all software has been approved for install and would be available in 15 minutes). But hey, what happened to sw XYZ? Oh that's right, company ABC (producers of XYZ) have an exclusive contract so that their software will be only available on service 2 because they have better terms than my service (service 1).

There's more, but this is long enough...

 

[darth maul]

No! This has been talked about by a lot of the mucky mucks in the puter world, Larry "can't look cool, even if in alaska" Allison has been a champion of this idea for a long time, and yet gets no one on his bandwagon. Ok, so we have this dummy terminal and a monitor right? Well where does my digital camera plug into? What about upgrading the video card every year? What about my sound blaster for playing games? What about game performance?!?!?!? What about a lot of the "extra" pci cards people have who's functions just can't be duplicated on a USB 2.0 or firewire port. Then you are going to tell the average user they have to pay a monthly fee to use software? Ya, ok, the average user won't even pay the extra $20 a month to go from 56k dial up, to DSL or cable modem access. Ok so you say, well the home user doesn't count, what about corperate users you say? Here in lies the only area this has and may countinue to preval in, although it seems like its dieing out in that areana as well. Sure there is a central server for need things, but it is still a real PC accessing it. Its just a bad idea.

Now if sound, video cards, and EVERYTHING else that you put in a PC stops getting faster or better maybe you can build dummy terminals that people would like to use.

 

[jbod]

There definitely could be some benefit to the centralized computing in the workplace today. Example: where I work, roughly half of all employees use their PC to connect to the mainframe. And that's all they do, all day. Turn on PC, open emulator, look up and enter data. All day. Seems like a terrible waste of resources to me.

On the other hand, the home user would not benefit unless like Elledan said and have a mainframe in their house. But wouldn't that be way too expensive? I'm not sure, maybe there is something that wouldn't be too hard to set up.

I personally feel that centralized computing will come back. But there will be a mix of both PC and dumb terms so a company can get the most out of their resources.

 

[Elledan]

<< I personally feel that centralized computing will come back. But there will be a mix of both PC and dumb terms so a company can get the most out of their resources. >>


Agreed.

 

[zephyrprime]

<< True, they won't be able to screw things up as many times as they do now, but what if their connection to the mainframe is cut off? >>



Users certainly won't be able to mess up their systems so much as they do now but the admins will be able to screw up the servers/mainframes. Servers were often going down where I worked.

 

[AcidzT]

im kinda divided on teh subject. i can picture teh house of the future with one central "computer" and a bunch of dumb terminals around the house (for teh geeks) and i can aslo picture things like thay are now. Sure your idea sounds good but whatabout when that new graphics processor comes out? will the central server have one for everyone that wants to use it? That stuff needs to be local.

Networks would have to improve alot. I jsut kinda picture it like how my non cable friends are trying to get on AOL on a friday night.

I wouldent "trust" my data out there, cause when tehy have to do matinence on wed night and i cant get my stuff im not going to be happy

 

[Shalmanese]

With the number of computers per person approaching double digits in some instances, I think the more pertantant question to ask is will there be a progression towards a cetralised mainframe Per person?

ie. everybody has a "god box" with the latest gadgets and doohickeys along with several thin clients scattered around.

 

[Train]

well, first of all, the indusrty isnt driven by the people who upgrade the thier video cards one per year or more, besides, this could be something done on the terminal side anyways.

The avg computer user uses 1/8th of his/her software, and the majority of that is Word, Excel, and Email, no need for a 2Ghz processor there. Dont you think that type of user would rather pay a fee of $20 a month to have terminal access to software and services versus paying $1500 or so for the avg computer? which is big and takes up space, and needs repairs, and patches, updates, etc.

Yes, i have actually heard of rational rose. and if the whole world goes back to mainframe, then each piece of software would be like a website, you go to the website and you start using it, they add the fee to your ISPs bill, or you purchase a one time license, or whatever. People here are too close minded, you say mainframe and they just cant comprehend how this could be good in the future, they are still stuck on video card speeds and storing personal information, pffft, think of the big picture guys.

 

[spidey07]

I'm not sure if this is a question of "will we go back to centralized computing". More of a question of how soon. PCs in an enterprise are a god awful waste of money and resources. Expensive, require WAY too much man power to support, constant upgrading (hardware, OS, software). The costs begin to scale even higher the bigger the company. UPS doesn't use PCs, almost exclusively windows terminals as do a suprisingly high percentage of large enterprises.

Think about it - would you want your company supporting 15000 computers and 200 servers or 15000 simple windows terminals and 210 servers? The capital costs alone on that one should make you raise an eyebrow. Then start looking at a 5 year cost of owership schedule - start throwing dozens of headcounts and contract labor spread all across the globe, cost of downtime, maintenance, Microsoft bending you over evey two years, extremely painful OS upgrades or rollouts. It'll boggle the mind. Now think 10 very large terminal servers (32 proc, 512 GB RAM) with a support staff of maybe 4 folks - sure makes sense. Upgrades done on a single maintenance window. No more performance problems with Sally's spreadsheet now! Have a problem with a winterm in Sydney? Tell your vendor to replace it within the guarnteed 4 hours the contract requires. Talk about good service levels. No more techs spending days fiddling with a laptop that doesn't act right.

Network speeds are so ludicrously fast today that servers can't even keep up. So yes, we have crossed the threshold so to speak where network communications are faster than the devices attached to them. Since a network is no longer a bottleneck it really doesn't make sense anymore to have a highly distributed environment.

I for one can't wait to get rid of PCs in the business. Nothing but trouble.



<< As networks get faster, here's my question: will we move back to a centralized system where our "personal computer" is just a terminal that doesn't actually do the computing, but sends and receives information from the central unit? >>


YES, it is happening today. Great topic BTW.

 

[CQuinn]

"Turn on PC, open emulator, look up and enter data.
All day. Seems like a terrible waste of resources to me. "

Speaking (somewhat) from experience, that was probably a much more
efficient way of doing it that was allowed by older mainframe technology.

The main benefit of tying PC in as "enhanced terminals" was that it
took a lot of the overhead or front-end processing away from the
mainframe, leaving it with more CPU time to dedicate to actually
processing the data that users would enter. And generic PC can
be made and maintained cheaper and more easily than custom-made
(and vendor supported) terminals.

-----

Train,

"The avg computer user uses 1/8th of his/her software"

I'd like to know where that figure comes from... and even if
it proves true, nobody can agree on the same 1/8th to use.

Centralized computing has a purpose, but the "big picture" has to
include uses that fall outside the range that is covered by the
use of a dumb terminal.

-----

spidey07,

The complaints you posted that boggle your mind now, are the same ones
that people were using in the 70s and 80s to move away from the high
cost of hardware and service contracts that companies were locked into
in order to have use of mainframe services.

"Network speeds are so ludicrously fast today that servers can't even keep up."

...If you set arbitrary limits on what the user can do on the network.
And who is paying / charging for this network bandwidth?

I would rather have a setup that meets both the needs and gives room for
the abilities of the company that is using the technology. Dumb terminals
might work for an Insurance agency (although my agent would probably disagree,
considering the number of times I've heard him gripe at the system while waiting
for some database lookup to come back). But those same "dumb" terminals would
be the death of a software development or web/graphical desigh firm.

We are all using a centralized data service (this forum) right now. But that
doesn't mean I wan't to give up all the other things my computer is doing/can do
for that access.

 

[MustPost]

<<The avg computer user uses 1/8th of his/her software, and the majority of that is Word, Excel, and Email, no need for a 2Ghz processor there. Dont you think that type of user would rather pay a fee of $20 a month to have terminal access to software and services versus paying $1500 or so for the avg computer? which is big and takes up space, and needs repairs, and patches, updates, etc.>>

If people were willing to not keep buying better comps, then why do people, even not techies still buy 1.5-2.0 GHz comps?
People think they need them and their not going to want to go back in time to a worse comp

 

[spidey07]

<< Network speeds are so ludicrously fast today that servers can't even keep up >>


My point being that no server is as fast as a well designed network. The people paying for the bandwidth (and also buying the net gear) are the same folks paying for the winterms.

My perspective is from a medium to large sized company where computing is mainly central anyways. A few heavy duty data centers with application/database/messaging servers. All other locations use that apps/data at these data centers. Doesn't make sense to use PCs in that model.

There are exceptions of course, but as a whole if you do the math centralized computing is significantly less expensive while at the same time providing better performance and less downtime.

 

[supernova87a]

I agree that if it were marketed correctly, this could take win a lot of the PC market. Imagine:

"A home PC for $30 per month, no hardware to mess around with, with free lightning fast internet access, all the applications and games you want, without waiting. Never upgrade, you'll always have the fastest processors, video cards, and latest software. Share music and movies with your friends instantly, and make free internet telephone calls to your heart's content, FREE!"

This is coming, I think. And you're right that businesses already know about the advantages.

And then, if we don't watch out for big corporations, they'll have us all.

 

[MustPost]

<<:"A home PC for $30 per month, no hardware to mess around with, with free lightning fast internet access, all the applications and games you want, without waiting. Never upgrade, you'll always have the fastest processors, video cards, and latest software. Share music and movies with your friends instantly, and make free internet telephone calls to your heart's content, FREE!" >>

What makes people think that the bandwidth to make this possible will exist. People have been predicting something like that could happen in the consumer market for years. They say bandwidth will soon be cheap enough, but it will not be in the near future. Think about everything that will need to be upgraded, routers, backbone, local telco conections. what happens if your network connection goes down all of a sudden, oops no more computers. Well many businesses can't function without their servers anyways so that might not be the greatest parellel.

It will be a long time, most likely never that long distance bandwidth will be cheaper than processing power. I understand that most of todays comps are sitting idle most of the time, but its still too expensive. The only way it would even be close to possible is if one or two megalarge corperations or the government banded together and made one huge system of shared computing, but that isn't going to happen.

 

[spidey07]

But it is happening. Bandwidth is so cheap you can buy performance.

Now if were talking home based computers, that would be tough to imagine.

 

[jbod]

Speaking (somewhat) from experience, that was probably a much more
efficient way of doing it that was allowed by older mainframe technology.

The main benefit of tying PC in as "enhanced terminals" was that it
took a lot of the overhead or front-end processing away from the
mainframe, leaving it with more CPU time to dedicate to actually
processing the data that users would enter. And generic PC can
be made and maintained cheaper and more easily than custom-made
(and vendor supported) terminals.

Speaking from experience here too, I can tell you straight up that the emulator does nothing but give you a GUI. Thats where all the processing comes in. At least where I work, the mainframe does 95% if not all the processing.

 

[mikey76]

This was Larry Roberts original envision (Driving force behind the ARPANET). Mainframe computing networked to dumb terminals. What he did not forsee was Smart Terminals at a reasonable price.

I am not sure that we aren't heading in that direction anyway. This forum for instance is a centralized computer networked to smart terminals. The bulk of the data is on computers not in people's homes.

MikeY

 

[Dragon23]

All can say is I hope that never happens. They may have been predicting it for the last 10 years. I know I would have to find a way to fight it. I like controlling who has access and who see's what I have.

 

[Zuidera]

bandwidth hmn ok now what about latency?

when I am playing Q3A / UT / RTCW and move my mouse the computer gets the message pretty quick over PS2 the program receives, reacts, the CPU & mem system crunch some numbers, the result is on my screen pretty damnde quick.

now please explain to me what it would cost to get my a *low latency* link to my friendly neighbourhood supercomputer?

cheers
Zuidera

 

[CTho9305]

well... at my university you can usually ping any server withing 7-20ms - pretty fast IMHO. you can run programs on them and set up an xserver on your machine if you are really bored; netscape runs fine, as do a fair number of other apps

I think with a few changes to broadband netwrks, this could be done over a larger area. of course, right now, its still way too expensive.

 

[Daovonnaex]

I do not think we will return to centralized computing (with exception to some large enterprises), because their would be fierce resistance from Tier 1 and Tier 2 PC vendors. The vast majority of their profits come from large corporate sales, and they won't simply allow their profits to vanish with the return of centralized computing.

 

[Turkey]

People just like having their own stuff... witness the number of people who own cars when buses may be cheaper/faster/more convenient, witness the number of people who still buy DVDs and CDs when DivX rips and mp3s are easily available with a little time, the number of people who buy books when they are available for free at the library... etc etc. Chances are if you have a leaf blower/snow blower/lawn mower/power edger, one of your neighbors does too, even though you could have easily bought it together and shared it with only minimal planning and a savings of hundreds of dollars. For whatever reason, people like having their own stuff.

As for cost, I think there is gross underestimation of cost going on here. The number of tech support people in a company is based on the hardware there, the software there, and the users. You can't just reduce the number of tech support people because you have less hardware... maybe it'll take less time to upgrade a certain piece of software than in a client/server model, but the software that people use is still going to be there, and people will still have the same problems as they did with the software in the client/server model. So the only reductions in cost would be a small percentage of tech support leaving because of reduced hardware and possibly overall less money spent on hardware. Going on the assumption that current hardware can't keep up with current networks (is it clear now why that statement doesn't make sense?), that means that specialized hardware must be used. Since there are very few companies that produce such hardware, there is a limited supply -> high cost.

So if you could get a high-end Dell wintel server for $3k that would serve x number of people at a specified level (used for any or all of web serving, file serving, mail serving, application serving, or something else) and you needed 180 production + 30 test servers of them to satisfy current demand, then to reduce that number to, say, 4 mainframes (3 production + 1 test) then each mainframe would have to support 60x users of each wintel server (understanding that mainframes have much higher processor/memory subsystem requirements than a server in a client/server model). Also, each would have a maximum cost of $160k just to break even with the hardware costs of the client/server model. And, service contracts would be overall *more* expensive than in the client/server model because there's fewer points of failure. That's because if one fails, your service level to users goes down dramatically as opposed to marginally, ie your mainframe vendor has to provide a higher level of service to you.

It's not like processing power is going to keep getting higher and software requirements or software usage will suddenly stop at its current levels. If you don't believe that software requirements will keep increasing after 30 years of continuously doing so, then stop and tell yourself: "Just because I can't think of another use for the power of this processor doesn't mean that any of the 5 billion other people on this earth can't either."

There will always be a place somewhere for all these computing models like mainframe/terminal, client/server, client processing, and peer-to-peer, and there may be isolated instances of mainframe/terminal computing replacing client/server in a corporation or a consumer mainframe/terminal service, but I don't think we'll see a mass comeback of the mainframe/terminal model.

 

[Train]

My moms hospital just switched to all Citrix Metaframes. Around 6,000 users. Big server with all the applications anyone uses and dummy terminals. Their I.S. dept is expected to save $4 Million this year.