Slashdot Mirror
Researchers Critique Today's Cloud Computing
Red Leader. writes "MAYA Design just released an excerpt from one of their forthcoming books as a white paper. The paper offers a different perspective on cloud computing. Their view is that cloud computing, as currently described, is not that far off from the sort of thinking that drove the economic downturn. In effect, both situations allowed radical experiments to be performed by gigantic, non-redundant entities (PDF). This is dangerous, and the paper argues that we should insist on decentralized, massively-parallel venues until we understand a domain very, very well. In the information economy, this means net equality, information liquidity, and radically distributed services (and that's pretty much the opposite of 'cloud computing' as described today). While there is still hope for computing in the cloud, it's hard not to wonder if short-term profits, a lack of architectural thinking about security and resilience, and long-term myopia aren't leading us in the wrong direction."
can someone tag "goodluckwiththat" please?
"...decentralized, massively-parallel venues until..." until its possible to slice, dice and securitize the cloud as investment opportunities. Oh wait...
I don't use the term at all. Putting thinks like network storage, Amazon's EC2, Google Docs and del.icio.us together is non-sensical. Yes, 'cloud computing' is a buzzword. I knew that. Maya's vision for a true P2P information network is nice though, albeit somewhat too idealistic.
Error 001
Security Scan and Virus Detection do not work with your operating system.
This makes no sense. Even the deduction makes no sense, in context. TSK TSK those idiots who invented the mouse were engaging in risky behavior?! Let's demonstrate insight by mentioning an economic trend that has nothing to do with technical innovation? Why would radical experiments be conducted by redundant entities? I am scared to download the PDF, for fear it's got more insight that will frustrate and elicit vitriol from me.
Often wrong but never in doubt.
I am Jack9.
Everyone knows me.
I totally agree. If you host your servers in one datacenter (cloud) you chose that one and not an other one. And if that datacenter brakes down, you have a problem. Same with clouds.
The article says a cloud works the same as a normal cloent-server application. On the otside that is true, but one dig difference is it should (if the application is desishould well) scale rather nicely.
Example: There is a website that compares different ideas of different parties and with a couple of questions it tells you that party is the closest to your idead (that is nice when you have 20 parties to choose from in the upcoming EU elections) But such an application probably has high traffic probably 2 months every four years. Then you pump up the number of servers and after the elections you reduce to one or two again. No costs of buying hardware and all the stuff.
Coming back to your point (and that of the paper): you pick one cloud to host your application and then why would you want to be able to communicate with different clouds???
Privacy is terrorism.
we allowed radical experiments to be performed by gigantic, non-redundant entities.
The Japanese call it "Hentai."
Schroedinger's Brexit: The UK is both in and out of the EU at the same time!
allowed radical experiments to be performed by gigantic, non-redundant entities (PDF). This is dangerous,
So we should break up the large banks, and replace them with an untold number of smaller, local banks that each follow their own strategy ?
Letting them go bankrupt should have exactly this effect. Destroy the whole, sell of the pieces one-by-one to the highest bidder.
And the alternative is propping up banks, running the extremely enormous risk that we've misidentified the cause of the current crisis ... (too much regulation ? too little ? Obama ? Bush ? Clinton ? CRA ? The devil ? Oil price ? Energy prices ? GW (not GW itself obviously, but the policies "to prevent it" are affecting the economy) ? I'm not arguing for anyone of them, I'm just saying there's probably good arguments for a lot of these factors)
If we misidentify the cause of the current failure, or fail to act on it, even slightly, then we'll have an even bigger disaster on our hands in a few months/a few years ...
So we should force the banks to follow a much more capitalist course, versus Obama's communist "fix" ... well one would have to admit that's a given.
(I'm using the capitalist/communist distinction in it's original distinction : centrally (government) directed versus distributed decision making. With these (long used) definitions Obama's actions are squarely in the communist camp).
This coming from a generation that still thinks web apps are cool.
Don't think of it as a flame---it's more like an argument that does 3d6 fire damage
With power outlets at nearly every seat on newer planes going from New York to Asia, cloud computing is great. The only problem is when they guy in front of you tilts his seat back too far. Of course, there is no redundancy, one laptop is expensive and heavy enough, thank you.
What puzzles me though is that the article tries to argue that on one hand the cloud concept is no different from client-server as it stands but on the other that the problem is the lack of interoperability.
A random Microsoft server can no more interoperate with a random Oracle or Apple server than a cloud service can so exactly how is it worse?
I also think the term cloud computing is just a bit jumbled. I think of it as the Amazon model, you basically design your server as a VM and then multiple copies of that get instanced as needed. The strange thing is that model is far more vendor neutral than anything currently on the market. In theory there is no reason why any company with the hardware resources can't fire up 1000 copies of that VM if you choose to change vendor. In effect cloud computing by that definition (which I believe is the most common) is no different than leasing servers from a hosting service at present, it just scales a lot easier if you need to.
The economic comparison is equally false. If for example Amazon were to oversell their hardware by 10% then all that happens is the sites they host end up running a bit slowly and people move off the service. The whole company doesn't end up in negative equity and going broke because of that. That metaphor just seems so wrong in this situation that it pretty much makes no sense.
If we were seeing a situation where web hosting and data center companies were merging wholesale while pursuing shaky business models then you could argue that there was a comparison but we are not. Cloud computing is a technical development, and until we see huge companies hosting the entire internet there is no real risk.
Slashdot: Proof that a million monkeys at a million typewriters can create a masterpiece
If you are a business trusting mission critical data or computing to the cloud, then you need to verify how well the cloud handles these issues. Just because they have a big data center doesn't mean that they have redundant services spread across the data center and multiple data centers in case one is hit by a disaster. http://www.datacenterknowledge.com/archives/2009/03/23/carbonite-lawsuit-reveals-data-loss/
Think Deeply.
Not it is not the same with clouds and that was the researchers point. The term "Cloud" should not be a marketing buzzword for scalable. There should not be multiple "Clouds". The entire point of the Cloud as they refer to it was that it was amorphous and ubiquitous. One of their nicest points is that you slap a brand name on it then it is no longer the Cloud.
The current crop of so-called "Cloud" services (like Amazon) are a decent enough attempt to provide this type of platform, but provides a highly redundant platform as a single point of failure is not the same thing as removing the single point of failure. Amazon's platform can still fall over (and did for several hours earlier this year). In real Cloud computing that would not be an issue because rather than being tied to Amazons servers / datacenter you could execute your code anywhere that provided the service.
One of the problems with the real Cloud of the 90s was that it became too successful. So now we don't see it anymore. The architecture of the lower levels of the internet is now firmly entrenched: tcp/ip, dns etc. That is the real Cloud, a platform that we can write code for that really is ubiquitous. Services like Amazon are a logical progression of that platform, but they are not the logical endpoint. When there is a standardised API for code that will run on any of the major providers, accessing storage from any of the major providers, and able to replicate across them at will (chasing the cheapest prices) then the Cloud will really have arrived.
Finally, just to answer your question in a more concrete and exact way:
It's important to drill through the marketing buzzwords that the paper is rallying against. By definition there are no multiple clouds. So really what you are asking is why should I want to talk to different providers within the same Cloud?
Resilience. Partly to avoid downtime, partly to avoid vendor lockin. If I develop my application for EC2 and then Amazon decide to get out of the scalable computing business - I'm screwed. It's a similar situation to the drm-locked media where the license servers shut down after the company stopped selling it.
Another reason - competition. If I can move my application between Google and Amazon at will then I will pay the cheapest price for my cycles. If I have to recode my application to redeploy it then it will require a huge price differential before I do that.
Slashdot: where don knuth is an idiot because he cant grasp the awesome power of php
The article is missing the point that many of the organizations that offer a 'cloud solution' (Amazon, Google, Joyent, etc) have already been experimenting with cloud computing for a long friggin' time, and the massive parallel experimentation phase was "who can grow without breaking". Now they're offering what they learned from that as a service.
A measure of the likely price impact of executing my information order?
Bullshit!
cloud computing, as currently described, Windows, is not that far off from the sort of thinking that drove the economic downturn
While slightly off topic, it's one of the more even and provoking "armchair" analysis'(?) that I've seen on the subject.
FanFictionRecs.net
This article is classic self promotion in the vein of:
1. choose concept
2. label evil
3. ???
4. profit!
At the risk of being constructive, if Maya actually spent time looking at what larger enterprises are planning over the next few years they would see the obvious architectural foundations for creating compute/storage/application pools emerging within enterprise data centres.
Extending these across physically or geographically dispersed platforms (eg DR) will be well within our technical capabilities in the timeframes usually discussed. To see these platforms interacting with external third party resources also seems suitable as a destination architecture regardless of, or perhaps because of, its distance. The positive effect such a common architectural goal has on the fragmented deployments I see all the time in Enterprise DCs, cant be a bad thing.
This wont be the only cloud outcome, especially for small and Internet scale, as the majors are massively committed to branded, propriety solutions but enterprise will do their thing, relying on enterprise technologies and this will be of far more import than some bodgie P2P concept from an enterprise perspective.
[This space is NotForRent]
So we should break up the large banks, and replace them with an untold number of smaller, local banks that each follow their own strategy ?
Letting them go bankrupt should have exactly this effect.
and throw out the baby with the bathwater.. e.g. the entire economy, the middle class way of life, and arguably the US as a stable nation-state.
The issue of systematic risk can be handled through intelligent regulation. Set a maximum net worth of a financially and managerially independent subsidiary.. this will make large corps more modular and less likely to suffer catastrophic failure without impacting ownership/imposing wealth caps.
So we should force the banks to follow a much more capitalist course, versus Obama's communist "fix" ... well one would have to admit that's a given.
Because free market capitalism worked so well right? *cough 1929 cough* *COUGH communism as an idea didn't just arise spontaneously, but out of very real social maladies from pure free markets COUGH*
VLC FOR MAC IS DYING! IF YOU DEVELOP, PLEASE SAVE IT!!
OK, sure, the "cloud" buzzword is annoying and not very useful. That happens a lot in our wonderful business. But saying that EC2, GoogleApps, and Azure are all dead ends because they're the products of large corporations is a lot of fuss over nothing.
No doubt the definition of "cloud computing" will evolve. For today, it primarily means not having to know any details about specific servers anymore, or worrying about how to connect to them. That's not a terribly original notion, but it is a big step forward.
To those of use who remember life before ubiquitous networking, ubiquitous data protocols, and ubiquitous storage, we are hugely grateful for what little bit of cloud computing we've got.
"We receive as friendly that which agrees with, we resist with dislike that which opposes us" - Faraday
Marketing guys should rather than talk about Smoke Computing for a real buzz.
Achille Talon
Hop!
So they are saying that the modern definition of "cloud computing" involves putting all you data in one place, so that it can be managed at a single point by dedicated IT professionals? Isn't that exactly what they used to call "mainframe computing"? Will we all be using 3270 terminals to access the cloud?
I've abandoned my search for truth; now I'm just looking for some useful delusions.
This is copied from my humble blog: http://thefortifiedhill.blogspot.com/2009/04/rejecting-cloud.html
To understand the reason the Cloud is a bad idea, we need to look at the short history of the web since the late 90's. The best example to look at is e-mail, but the same arguments apply to most Cloud applications. In those days, you got email access through POP and later IMAP. The service you were paying for was just a reliable email server and an account on it. Some free sites generated ad revenue by injecting text ads into the bottom of your emails (some still do). Since then, various web clients have come about. The main reason for the exodus to web clients was not that the web email clients had a better user-interface but because users could have the same user interface at any computer terminal. This is incredibly attractive to many users who do not care that much about their email client's features.
There are a number of problems with "Cloud" applications.
First, is that you cannot access the service unless you're online. You may argue that you're almost always online, but this is hardly true for mobile users. The fact is that a user should be able to access their data even when offline. This is the reason Google has their Gears project. There is another implication: you don't posses and own the data. You don't actually know if it's safe, being sold, or even managed securely. You just have to take it on faith that the service provider is doing their job. In some cases you may not even be able to extract your work or data in a usable way. What if the provider goes under, what if they are bought and the service is canceled? Not completely owning and possessing your work should be a major concern for users.
Second, the user gets an augmented version of the service. By this I mean that the service starts to be branded and enhanced by the provider. It gets integrated with the providers other features and starts delivering unique features. At first this seems like a good thing, a result of competition between providers. The reality is that this results in broken APIs and interoperability issues. These enhancements make it difficult for a user to leave a provider or to integrate a provider's features with their other work.
Third, the browser is a bad platform for these kinds of applications. The browser was never designed to be a host for dynamic applications of this complexity. Their are numerous development and usability issues in web development. Almost all web work involves hacks and workarounds to accommodate situations where browsers don't adhere to the web standards. The browser has been contorted to fill a role that your computer environment should have filled all along.
I've just been reading overviews of the ZFS, and this white paper uses a lot of the same buzzwords with what appear to be the same meaning. ZFS is all about reliability, scalability and freedom from lock-in to any device; just like the cloud. I understand cloud computing to have four components:
1. a single ZFS pool which consists of every storage device in the world, using open-format data that are accessible to anyone, anywhere with proper access, and not available to anyone, anywhere who doesn't.
2. applications flexible enough to run on any processor in the world, as in the early descriptions of JAVA with its "write once/run anywhere" theme. In this ideal world, anyone could sit down at any machine and find their data available to them from this one pool, and their apps would run on whatever processor is available from that one pool.
A side effect of this portability is that data would always be migrating to the cheapest provider, and CPU cycles the same. There is no room for vendors to raise prices, as all their customers would instantly and automatically migrate away from them.
3. a reliable way to make micro-payments for those services. Literally in millionths of a cent.
4. reliable security of the data regardless of where it physically resides. One image that I can't shake is what happens in this ideal world when the Chinese government opens a huge new storage farm with a dedicated hydro-electric dam. Suddenly they have physical access to every bit in the world.
The ideal is virtually unlimited, virtually free computing for everyone. Service providers hate that. They want to be able to set their own prices for their locked-in customers by making it cheaper to stay and pay the higher rates than migrate because migration costs more than they'd save. So far, what we are seeing as the clouds are moving in the opposite direction from the cloud.
People should not fear their government. Governments should fear their people.
Google's Gmail interface is one side of a P2P.
Cloud metaphors in place of FUD, blue smoke and mirrors for purposes which remain, well, clouded.
What? Of course those aren't leading us in the wrong direction --- they aren't LEADING anywhere! Science still LEADS technology development, and Slashdot has even run multiple stories about the Open Science Grid. OSG, of course, is just ONE huge example of a set of massively distributed parallel services, spanning multiple organizations and institutions across the country, and peering with other grids across the world! Funding for grid development is completely unmatched in computing projects, given that it happens in so damn many places, and yet the need for faster and better performance (with a complete disregard for profits) drives scientists to find the cheapest and most efficient means of squeezing performance from hardware. It's a VERY active field of research (I happen to work in it)! The grid doesn't only offer compute cycles, it has storage and other services too, and cross-compatibility is a major concern. It really all MUST work together in order to be effective!
As scientists iron out bugs with this stuff, and as their students leave into the workforce, the technology slowly leaks into companies, most of which move at molasses-speed to adopt new ideas. Seriously, when *I* left this field 5 years ago, no one knew what the hell I was talking about when I talked at interviews about my work in "distributed computing" or "parallel processing", forget mentioning "grid" or "clouds". I gave up on jobs after a couple of years and came back to research, but it was just THIS PAST YEAR that any major companies noticed those catch words on my resume. Seriously, it takes a long while for them to catch up to development.
So, in short, it isn't myopia of thinking that's making cloud computing drive itself to the ground, it's the myopia of this article that makes it look that way.
-=[You cannot consistently judge this statement to be true.]=-
I have said this before, and I will say it again, this is EXACTLY what Stallman meant to say on the topic, he simply could not articulate it in a rigorous manner, but, that does not invalidate his point in the least when he said "it's a trap"
I agree wholeheartedly. I use SOME "cloud" apps, and use them in a calculated and limited manner, but, I will resist having my computing resources under the control of a bunch of filthy Suits someplace (because, ultimately, the major "cloud" apps ARE and/or will be under the control of some Suit swine)
SARAVA!
I've had my first true cloud experiment a good week ago. It saved the sites of dozens of customers hosted at that same server.
Without clouds, bandwidth & servers had to be upgraded within 5 hours to withstand the load. With clouds, all load was handled perfectly by putting the highest load (images and javascript) on the cloud and leave the pure webserving to the server. It brought the load from unworkable to fully bearable. I've written about this at my blog.
This new form of clouding could for sure hurt hosting providers at the end, since they will be moving less data because of cheap clouding. I just can't see why it would be bad to get the prices of hosting and bandwidth lower because of new technologies. Let's hope the ISP's won't make the same mistake as the RIAA and others are doing by not thinking/ignoring digital.
--- I am known for the ones who want to find me on the net. Is that a privacy risk or a privilege? One might wonder..
we should insist on decentralized, massively-parallel venues
We did. It has worked astoundingly well. The centralized, massively-serial venue was called the mainframe. The decentralized, massively-parallel venue is called the desktop PC connected to others via LANs and the Internet. Isn't the paper preaching 'do more of the same'?
I would argue that the 'until we understand it very very well' part hasn't come to pass yet... We haven't finished understanding network effects, or Youtube wouldn't be losing money. If Google, king of the leveraged network effect, can't figure out how to sustain it, I'd say we still have a ways to go.