Any recommendations for good articles and papers on cloud computing?
April 4, 2011 9:44 AM   Subscribe

I am interested in cloud computing and I am looking for recommendations re interesting and worthy articles and papers on the subject of cloud computing. One area of interest is in how it will change the old-fashioned client-server network model - the status quo. Is outsourcing the network infrastructure really going to produce a dramatic transformation or is it somewhat over-hyped and exagerated.
posted by conrad101 to Computers & Internet (12 answers total) 12 users marked this as a favorite
It won't change the client-server model. The server is just running on somebody else's hardware. P2P threatened to change the client-server model, but has fallen out of favor for a number of good reasons.

Cloud computing is more of an economic revolution than a technological one. The one big advantage it offers is elasticity --- if I'm running a web startup on the cloud and I get Slashdotted, I can pay some money and immediately get the capacity to handle it.
posted by qxntpqbbbqxl at 10:09 AM on April 4, 2011

To my mind, "cloud computing" is just the latest argument in the ever-ongoing thick-client/thin-client debate.

One area of interest is in how it will change the old-fashioned client-server network model

It doesn't. At best, it just moves the distinction around in the chain.

But, from a programming perspective, it makes absolutely no difference.

Is outsourcing the network infrastructure really going to produce a dramatic transformation or is it somewhat over-hyped and exagerated.

It is definitely over-hyped and exaggerated. No freakin' doubt about it. And we'll see shit shift back in the opposite direction just as soon as we have a couple major disasters (either technological or legal). Just like it always does in these situations.

The only thing cloud computing really offers is real time hardware scalability. If you design your software correctly, you can add and subtract capacity in real time based on instantaneous demand.

In practice, however, very few shops actually design their software that well.
posted by Netzapper at 10:14 AM on April 4, 2011 [2 favorites]

I'm interested in the topic of cloud computing from a research perspective, but by no means an expert in it.

One article I found really interesting was Tim O'Reilly's The State of the Internet Operating System (Parts 1 and 2). O'Reilly talks about the major players, the services, the connection to mobile computing, and what's still missing.

There is also a good technical overview of cloud computing services on the Communications of the ACM website called A View of Cloud Computing, talking about service levels, programming APIs, data centers, the economics, and major challenges.

From my perspective in doing a startup, cloud computing has dramatically made our lives easier in terms of getting computation and storage capacity when we actually need it, instead of buying (and then having to administer) new servers. The collection of services that Amazon offers is also quite impressive, lowering the upfront cost of doing business for startups.
posted by jasonhong at 10:25 AM on April 4, 2011

One of the best things you can read about deploying on the cloud is Netflix's post-mortem on what they learned in moving from their own infrastructure to Amazon:

The cloud has great potential - IF people build their software and their overall systems to be work - and more importantly not work - in a co-tenant environment. You're going from dedicated resources to sharing resources with who knows what.

People generally aren't doing that or even talking about it at this point. You'll be much further ahead of the game of you search the experiences of people/companies that have actually deployed revenue generating systems on the cloud and lived to tell about it.
posted by NoRelationToLea at 10:46 AM on April 4, 2011 [1 favorite]

The only thing cloud computing really offers is real time hardware scalability

Yes but that's the whole point!

The CACM paper referenced above is good, and there is a longer version available as well.
posted by sesquipedalian at 11:02 AM on April 4, 2011

Response by poster: Thanks everyone for the responses! I've had a look at a paper called: "Above the clouds: A Berkley view of cloud computing" and the picture it paints for cloud computing is very impressive. It sounds a big deal, treating your network like it was an essential utility service like for example electric. At face value atleast it builds a strong case.

On the opposite side, some like Larry Elison makes out and that it isn't new, it's just a fancy marketing term for stuff that we have been doing for years.
posted by conrad101 at 11:08 AM on April 4, 2011

Yes, little in cloud computing is "new". When I took systems classes at Berkeley, some folks there said that none of the ideas were really new, it was just applying them in new contexts or at different scales.
posted by jasonhong at 11:26 AM on April 4, 2011

It sounds a big deal, treating your network like it was an essential utility service like for example electric.

It is a big deal, but more for economic than technical reasons. Cloud computing doesn't replace processors and routers, it just gives a framework for tuning the number of processors we allocate to given problems (like serving web pages). The idea of utility computing isn't new to anyone who's ever paid hosting fees. Ellison's right though, "cloud" is a marketing term intended to divorce people from preconceptions about "grid computing" and whatever market failures people remember.

What cloud vendors like Amazon have done is make the initial investment in a supercomputing facility to rent out surprisingly cheaply to people. If you think about computers in terms of problems solved per dollar, and you've only got one problem you want solved by tomorrow, building your own cluster looks like a bad deal vs renting out someone else's. For a certain category of problems, we now have a financial tradeoff that was previously quite risky. Amazon's job is to eliminate the implementation risk that building and designing a huge datacenter and compute cluster entails for many folks.

But none of this means you'll be building a peer to peer cloud. Firstly, amazon has no interest in helping customers migrate servers away from their datacenter. No vendor does, since their investment can't compete with an investment in faster cheaper hardware a year from now. They'd lose their customers rapidly without some friction. Secondly, part of the benefit is Amazon hires the smart IT people to build the infrastructure and distributes their skills to far more clients than otherwise possible. So if you were hoping to build a small datacenter in your basement and rent it out, well, cloud isn't usually that.

Personally, I think cloud storage is far more relevant to individuals than cloud computing, and the P2P story is worse there. With computing, your demand is bursty and subject to peak loads that are uncorrelated across systems (or even negatively correlated. stock quotes during the day and football scores at night, for example)., and algorithms can reallocate. In contrast we generally want data forever. Cloud storage is useful for increasing redundancy, but I don't know many who place their full faith in it. So there's not really any fundamental trade to be made.

Anyways, I've gone on far too long without sufficient citations. Fundamentally, cloud isn't new, but the fact that an online retailer, who doesn't have existing IT sales and consulting margins to worry about is.
posted by pwnguin at 12:23 PM on April 4, 2011 [1 favorite]

A lot of the big-wig iPad-toting execs at the large financial firm I work for are very hot on the whole "cloud" concept. But then when we tell them what it really means, they're like "Oh, we're not really that keen on giving all our corporate emails and financial data to a third party. How about a 'private cloud'?"

So yeah, we'll start referring to the data centre we've been running for decades as a "private cloud". Can I have my bonus now?

Not bitter, much.
posted by Diag at 1:02 PM on April 4, 2011 [1 favorite]

"Cloud" is a game changer for IT. Screw the over-sold hype, the big fact is this - Cloud means IT departments don't have to worry about infrastructure like we do now. Not server uptime, not patching, not hardware upgrades, not networking, not cabling, not cooling, not power, not redundancy, not physical security, not upstream connections, not hypervisor settings, nothing. The "server" in client/server is now outside of your reach and concern. You pay X for Y service.

Ponder that for a moment - what would your IT staff do if they didn't have to touch server hardware?

Sure that underlying concept isn't new, but the high bandwidth WAN links and VM computing we have now is. So what we can leverage 'the cloud' concept for is greatly enhanced.

how it will change the old-fashioned client-server network model - the status quo

Most companies are in the business of things other than running a datacenter. Cloud means they can get out of the datacenter business and focus on the core business (but for real this time) where that makes sense.

example: If I'm CompanyA planning a major upgrade to my mail system, Do I as a CIO spend my limited IT budget on a typical upgrade path (servers, OS and app licensing, personnel costs) or do I put it in cloud instances and services where my dollars go to other items like SMEs, user training and support staff? Do I assign my limited staff to deal with hardware support issues like firmware upgrades and OS patching, or would I rather them focus on managing the services they're providing to the company?

And that's just the 'server' side. Cloud VDI is coming sooner rather than later. Hosted apps are here today - If I'm an IT exec, I'm going to ask my staff some hard questions whether installing instances of MS office (for example) on individual computers is worth the deployment costs when we can stream the same experience from a few servers and not have to touch anything on the client system or worry about upgrading PC hardware.

There are pluses and minuses to each solution, but simply being able to answer the questions means IT staff and management need to understand the concepts of cloud based systems and IAS. Like "Can I legally put my data in the cloud?" Some people have hippa/sox/other requirements that prevent any disclosure to 3rd parties or limit to cloud servers in US-only data centers.

Putting on my crystal ball hat, I'd say that inside 10 years (2-3 upgrade cycles) most business will be doing all or most all of their IT in some form of public cloud or public/private hybrid. the economics and benefits are just too big to ignore. And that's the real reason for the cloud hype. This segment is going to be huge, and every major IT player is moving there in a big way.
posted by anti social order at 2:39 PM on April 4, 2011 [1 favorite]

« Older Broken plugs from an RCA cable are stuck in my amp...   |   ‚ÄúThose who are clever, who have a Brain, never... Newer »
This thread is closed to new comments.