Anyone have any experience with cloud computing?

Any user experience? Serious supercomputing stuff like university research or software development; or more 'non-serious' stuff like Google Apps?

Or maybe you work with one of the providers? Amazon, Google, IBM, Microsoft, etc?

What do you think? The next big thing? The last old thing come round again?
Will cloud computing fly? Or are security/privacy issues too much to overcome?

Comments much appreciated. Thanks.

p.s Mods: if you feel like this belongs in OTB, thanks in advance.

Didn't I respond to this thread already?

Hmmm...

CloudComputingExpo has some great, well, expos. Definately worth your while to go press flesh with all the leading contenders in the business.

Originally, it was to be similar to project SETI, in that it would make use of underutilized processor, RAM, and HD storage space, by means of a network operating system that literally ran on all the networked computers, rather than as a server on one of the networked computers. Thus, you would no longer need a domain server, file server, print servers, and application servers, or server farms, as the entire pooled collection of computers became the server farm. The NOS would supposedly load balance, on the fly, creating the best mix of localized data and distributed HD storage. As RAM and processor speeds are so critical, while localized data storage is far less critical with high-speed lans like 1 and 10 Gigabit ethernet, data-intense operations would be the most localized, while the data itself was to be distributed in a RAID 6-like format, where each computer in a subnet would act like a different drive, and the two parity volumes would be attempted to be put on two different network subnets, one in the same subnet as the data (but different computer) and the other in a different subnet.

Supposedly, there was enough redundancy that one could turn off just under 33% of all computers and assuming the computers were fairly equal in terms of capabilities, the system would chug along as if nothing had happened. When the other computers booted, the system would fix itself, bringing those computers back up to speed and online.

Later, the term cloud computing came to include browser/web-based services that were stored on either intranet or Internet servers. Indeed, most of these servers do embody the original view of cloud computing, but as a virtual server for relatively thin clients.

My personal belief is that it's already morphed into the most viable format:

Relatively thin clients, accessing apps and content stored in the cloud. These can be heavily managed or just POSBC (plain, old, store-bought computers) owned by employees. The security is much like those used in banks - very heavy, but it's all taken care of at the server, and secured via SSL to the clients, so it matters little as to what's on the client (assuming an honest client), as the client sees but a tiny portion of the information. Good hardware/software approaches at the client will further secure the environment.

Cloud of computers linked together, sometimes as a grid, which store the data and apps, and serve both, while performing the client billing functions.

I've not had much experience, except for setting up ICE sites for the DoD. It's kinda-sorta cloud type stuff, but not really, as it's mostly link/document based, rather than app/data based. Furthermore, the actual collaborative part hasn't taken off, as the users' mindsets appear locked into hierarchical thinking, rather than collaborative involvement, and that mindset is reflected in the way they develop their ICE content.

It's a *bit* frustrating seeing even bright people commit the same mistakes over, and over, and over, and over, and...

I'm afraid that except in a few rare circumstances, the way computers are used today remains little more than as a glorified typwriter. E-mail is nice, but that's still a text-based system, rather than a database system.

How often have you received an e-mail at 9:53 saying, "Don't forget! There's a mandatory meeting at 10:00!" and know for a fact it's the first time the word went out... Even using Exchange's calendar, and inviting all members of the department (or just the ones selected for the meeting) would be a step above the way things are done now.

__________________
If I set the budget, we'd have Ares and more. Unfortunately, I don't set the budget, and Ares is just too expensive and too far out for us to accomplish our goals within the budget we were given.

If we halt the ISS, all versions of Ares, and transport Orion and Altair aboard DIRECTv3's Jupiter family of Shuttle-Derived Launch Vehicles, we just might make it back to the Moon by 2020.

Thanks Mugs.

The concept isn't new, but finding a way to make it scale well has been the issue. It's also not easy to make it work in a hetrogeneous environment.

Thanks Halcycon! I've read that scalability is one of it's selling points. Is this right in your view? Or is it still an issue?
Last amateur techie question: what's a heterogeneous environment?

Two different types of computers. Or more

Ahh...

Like a definition...

__________________
*

One of the first articles I saw was this, 'Twenty-One Experts Define Cloud Computing'.
If you need twenty-one experts just to define it...

I looked at clouds from both sides now,
from up and down, but still, somehow;
Its clouds illusions I recall;
I really don't know clouds, at all.

__________________
At night the stars put on a show for free (Carole King)

One Earth, One Sky - IYA 2009
All moderation in purple

Cloud Computing has a variety of meanings as others have mentioned. Some of them are not very clear, and being involved in this sort of Environment, it's simply a catch phrase that CIO's like to use to impress, and sometimes beguile people.

(OK, so I'm a Cynic)

There are better terms to be using depending on what is going on.

Grid Computing: Many computers/platforms that have centralized datastores, and can shuttle processes (running applications or servers) around on the grid, or dynamically apply more of the grid to a process. Basically the ultimate in Client/server where the client is usualy a browser, or some other thin client. Some examples of this are IMB's in a sysplex, Blade Towers hosting VM's, etc....

Distributed Computing: This is more an Application centric platform, where the design utilizes recources on a variaty of platforms, including using recources on the Client computers. Examples would be MMOG's, or GUI applications with server backends.

SETI would be a good example of an Environment that is both a Grid, and Distributed.

__________________
There is no problem that cannot be solved by a suitable application of high explosives - US Army Demolitions School

I just saw Hayley's comet, she waved, Said "why you always running in place? Even the man in the moon disappeared, Somewhere in the stratosphere" - Shinedown

http://worldsofothersuns.home.comcast.net/

Thanks dgavin. I'm a cynic too. The thing is, from what I've gathered so far, cloud computing is being pitched by tech firms as a way to get rid of client companies' CIOs, or at the very least massively reduce their budgets (corporate death!). Have I got this wrong?

Shucks, I thought it had something to do with clouds, which might lead me into related nebula work (maybe).

Let me know when the 9th version is out.

__________________
Lighten up! This is a stellar board!

I'm confused. Is 'cloud' an analog to 'distributed' ala BOINC?

__________________
---"Why do things have to suck so bad?" a friend once asked me.
"Because space is a vacuum and that's a lot of suck." I replied.
(Actual quote)---

I wouldn't worry Lotus. As far as I can see, everyone's confused, including techies.

Cloud computing includes, but is not limited to, distributed architecture as in the BOINC sense. In fact, 'cloud computing' seems to be more of a functional description (rented services) rather than a physical (architecture) one. But I'm not 100% sure- I'm a little confused too. Hence this thread. Also, as George says, a lot of this is old models re-hyped and re-packaged.

Intro stuff, if you're interested:

Wiki: Cloud Computing
BusinessWeek: Computing Heads for the Clouds

I tend to see cloud computing as "mapreduce in a heterogeneous network", mapreduce being an algorithm schema, a particular way of organizing computations. Some might mean it more generally than that, but practically, implementations of cloud computing tend to use some variant of Google's original MapReduce algorithms.


Google MapReduce (and I mean google as a verb this time) for more, including Google's own papers on the topic.

__________________
-----
Todd (Bowie, MD, US, North America, Earth, Sol System, Vega region, Local Bubble, Orion arm, Milky Way Galaxy, Local Group, Virgo A Cluster, Virgo supercluster, the universe in which spock is clean shaven)

Quidquid latine dictum sit, altum sonatur.

personal page: http://blog.astrosketches.info

Hey thanks tdvance. That's a good branch to follow.

Basically, from what I've seen, you about have hit the hammer on the nail.

Usualy the people pitching 'Cloud Computing' are salesman that are repackaging some older terms 'Datacenter Consolidation' and 'Downsizing', and trying to wrap it in with distributed computing and outsourcing models.

And considering the state I work for is in the middle of one of these 'Datacenter Consolidations' (and it's about the 4th one I've been through during my career) I can say, honestly, that it's all just a bunch of smoke and mirrors. To quote a favorive TV series, "It's Felgercarb!"

It never works, it ends up costing more then it was supposed to save, and if you go with full outsourcing, just getting a copy of your data back from the vendor usualy costs millions of dollars, and they always reformat it before sending it to be almost as unusable as possible. (Yes E.D.S. I -am- talking about you here, how did you know)

And, of course, they never listen to thier currently employeed IT staff when they go ahead and do it, no matter how much you point out what is wrong with thier planning. Or even if you give them a proper plan. There is a proper way to consolidate datacenters, but I have yet to hear of anyone actually doing it right.

__________________
There is no problem that cannot be solved by a suitable application of high explosives - US Army Demolitions School

I just saw Hayley's comet, she waved, Said "why you always running in place? Even the man in the moon disappeared, Somewhere in the stratosphere" - Shinedown

http://worldsofothersuns.home.comcast.net/

I just got back from Microsoft PDC, so I've had about as much cloud computing discussion as I can take for one lifetime. Aside from a nice new Windows Azure backpack, I did not find much to impress me about cloud computing. Microsoft has done a nice job cleaning up the APIs for all of their services, but because Azure is tied to their own server farms I don't think they are going to revolutionize cloud computing. I could not find anyone in their Azure "experts" area that could answer a simple question: "Why should I tie my development to Azure APIs when I am tied to a single service provider who won't even tell me the price structure until I'm too far along in development to change my mind anyway?" Perhaps I was a little to snarky, as nobody I asked that question gave me free T-shirts.

To be fair, Amazon was there and didn't fare much better with the same question. They at least had the advantage of already being in business so they could give me real prices. Oh, and a free T-Shirt. Their services are not quite as broad as the Azure platform--not that they necessarily need to be.

I've already thought of cases where I can use cloud-based SQL and other services to get around having my own IT department set up a server--which can take years and is simply not worth the frustration for an internal productivity app.

__________________
Do try not to take me too seriously.

Thanks guys. Perfect- just what I was looking for.

Once tried to deal with a subsidiary of theirs, MphasiS. I wanted a simple economics database, but after 4 meetings with a guy in a really expensive suit and getting nowhere, I went and bought off the shelf.
Actually, to be fair,that was just me being stupid and inexperienced. I just wanted something simple, and I should never have gone to MphasiS in the first place.

Ah, so you went to the conference did you? Had a good time I hope.
I find the lack of price information bizarre to say the least. It's supposed to be one of the main selling points! Nutty.
By the way, you're not impressed with cloud computing...but you're thinking to use the services anyway? This is my fault perhaps, not getting your phrase 'cloud-based SQL'...

Thanks again you two. 'Almost, but not quite there' is the impression I'm getting...

Well, what I mean is I cannot see using cloud services in the real products my company works on--lack of internet connectivity at customer sites kills the idea pretty quickly. I'm not seeing sufficient benefit from cloud services to overcome what I see as the major weaknesses:
1) Dependency on a 3rd party to provide a business-critical function. I don't expect Microsoft or Amazon to go out of business per se, but suppose they decide the particular service I'm using is non-profitable, and phase it out?
2) The "moving target" development platform. With traditional development tools, you can usually count on a stable platform because you are completely controlling its deployment. With a cloud service, they can make a "minor" tweak to an API and force you to spend a lot of money fixing a "finished" product.
3) Uncertainty of internet bandwidth. I found it highly amusing that Microsoft was pitching online versions of Word and Excel at PDC, yet my Wi-Fi connection in the audience at that same presentation was unusable. Good thing I was taking notes with a good-old traditional run-local application.

That said, I can see cloud services being very useful for small companies and small, short-cycle development. If you are just putting together an app that tracks your company's Christmas Basket donations for 2008, it doesn't make sense to put together your own SQL server for it. It is very convenient to buy some space on Microsoft's SQL server in the cloud for a few months, and use the very slick development tools on the Azure platform to make writing the app for it very easy.

__________________
Do try not to take me too seriously.

Good points Demi. I can give you a client's perspective. I run a small financial company, and as far as cloud computing goes, security and confidentiality are my biggest worries. Not my data you understand; our industry is pretty good when it comes to security, and we're all pretty used to storing our data on servers that belong to other financial companies. But my two problems:

1. Trusting a non-financial company with my data. (It matters)
2. Trusting another company with my clients' data. (Liability)

Having said that, we do generate a ton of data that isn't so confidential, and dare I say useless. Would be nice if I could shift some of that one day.

Hi, PraedSt. I can't help you much further, but we did use such a system at ESA. You could assign a task to the network, it would be given a priority, start time and I believe also a capacity request, and it would then use I think both spare capacity in the network and capacity from dedicated crunch computers. I'm not sure on the details, as I haven't used it. I was the guy making analytical models, that could run on a single desktop in a few seconds. It were the numerical boys that needed these nights of computer crunching. ha!

btw, I misunderstood your question in PM completely. I had nearly linked you to the MSc thesis research of a friend of mine, who studied 3D computer cloud generation.

__________________
To the regular visitor of internet bulletin boards it is clear that it's an excellent idea your parents get to choose your real name.

Lol. Thanks. ESA though- WOW! Very impressive.

And then you don't even know what I worked on at ESA.

__________________
To the regular visitor of internet bulletin boards it is clear that it's an excellent idea your parents get to choose your real name.

If you have to kill me, don't tell me.

I worked in the R&D department, on scramjets. Basically, I worked on the design of a civil, upscaled version of what is in popular media the military hypersonic concept plane "Aurora". This project was/is the lesser known big brother of the Lapcat M5 scramjet project, which seemed a bit media-addicted. Bigger and faster than that Popular Science thingy. I've seen other things pass too in our group, but if I told you about them, I'd have to kill you indeed.

I made an analytical model of the scramjet engine, useful for making fast estimates and starting points for the numerical guys. Quite a new concept for mathematical models of this scale. Not everyone liked it that when they shouted an idea at a meeting, I could give a very good estimate of the resulting performance and shock patterns within seconds. They were used to having to wait weeks before being proved wrong . The analytical model is also a good aid in designing inlet geometry and geometry transitions. What we called, the sculpturing. One of the big arts of scramjet design.

The best part was that my model didn't need cloud computing or anything. It ran on a simple Dell desktop, easily. Didn't even need a workstation for it. Of course, ultimately it was less precise than the numerical calculations, and for detail level things such as injection and mixing of fuel simulation, it was not suitable. But on the positive side, it could give precise answers on overall shock patterns and more than performance answers on performance (for these cases where combustion was not too chaotic or bad) within seconds rather than hours or days.

At the speeds and temperatures where we were "playing" with our engines, you can leave reference books behind. You're on your own there, often going where nobody has gone before. In the best case somebody else also knew something about what you were doing. But reference books were used up to their limits quite soon indeed. And all kinds of nastiness happen to gasses when you torture them like that.

__________________
To the regular visitor of internet bulletin boards it is clear that it's an excellent idea your parents get to choose your real name.