How is the "cloud" any different from the "net" (as in Net Computer, or NC), which he promoted so heavily in the mid nineties?
"That magic part would connect the Network Computer to the Net. There would only be rudimentary software and memory on the Network Computer. Most software and serious memory would be out there on the Net where it could be easily maintained. The system would run on Java and use Oracle databases. Microsoft software would be nowhere in sight."
I think that's the point he's trying to make. He's basically just saying "buzz words suck" and then riffing on the term Cloud Computing for a few minutes. It only seems contradictory because Techcrunch framed it as commentary on cloud computing rather than the silly joking around it's supposed to be.
In fact, Techcrunch actually misquotes him. The Techcrunch article opens with "According to Larry Ellison, it’s nonsense and water vapor" when in fact what Ellison says is "[The Cloud] is NOT water vapor" (the point he was making is that the cloud is just actual computers on the other side of a network not "a cloud" which is where water vapor comes in)
Oh for the love of god, Larry, shut your smug mouth.
It's so easy to sit back on your mega yacht and dismiss cloud as something that's always been around and always will be and blah blah blah, but for a little startup it's MANA FROM HEAVEN.
To be able to do massively compute intensive or storage intensive tasks on demand without sinking a million dollars of capital into a datacenter is the difference between success and failure for my little startup.
How many times have you heard "oh, there's a ten week lead time on that server" or "there's no more power in the cage" or "there are no more cages at the datacenter" or "i'll be up all night recovering the raid array" or any number of other hidden costs in owning or trying to own hardware?
We check our cost projections for buy vs cloud almost every week, and when the time comes we'll definitely buy, but for we couldn't be doing what we're doing without it. Sure it's overhyped, but so was the fucking web. Yes, there's weird performance stuff and not always enough boxes, but so what? Sorry if I'm missing a subtler message he had, but i had to turn off the video after about 30s because it hurt my brain.
He's not criticizing cloud computing, he's saying "cloud computing" is an idiotic idiom. And it is. "Cloud computing" is just distributed computing. It has been around for decades. (See: http://en.wikipedia.org/wiki/Distributed_computing)
But, it's the latest tech word-soup. Like in 2005 when everything Javascript became "Ajax", everything distributed becomes "Cloud".
As far as idioms go, it's as good as any. "Cloud computing" != "disributed computing" -- it's on-demand virtual hardware and storage. That's how most users of cloud think of it, in my experience.
I was hugely skeptical of the cloud concept -- sneered at it, even -- until i started doing the cost/benefit analysis.
On-demand virtual hardware and storage is Infrastructure as a Service. I'd be willing to bet that most developers think of that when they think of the cloud (since, as you point out, they're actual users of the cloud).
But then some marketer somewhere (I'm guessing Salesforce, but that's my own bias :) decided that any web application could call itself "Cloud" (since "Software as a Service" was no longer cool. Again, I'm blaming Salesforce for flogging that horse to death...). This way, even normal people can use the cloud/buy cloud services, and it's just as easy as using the Web since it's really just the Web! (Way to go, smart and savvy Internet^H^H^H^H^H^H^HCloud user—you're so much smarter than everyone else and can expect to get promoted over/laid more than your coworkers still using that ancient Web 2.0!)
The Wikipedia article on Cloud Computing is a good illustration on how vague the term has gotten, especially how ill-defined all of the sub-classes are: Platform as a Service, Application as a Service, Service as a Service. Pretty much anything on the Web can be made into a definition of "cloud" somehow.
I'd really like to see Cloud Computing only applied to things involving immense elastically scalable computing and storage that can be provisioned and taken offline immediately, but I don't think that's going to happen. The meaningfulness of the term has been destroyed by the rampant "me-too"-ing of the industry.
The field developed and it became useful to distinguish between different uses of Javascript, so new names were invented to do that.
Javascript that changes the page and calculates on user input is still 'Javascript'. Javascript that sends requests to a server and updates the page based on the result became 'ajax', Javascript that does processing in a small-tool-page-independant fashion became 'bookmarklets', Javascript that does page processing in an after-market-modification fashion became 'userscripts'.
Likewise distributed computing - that just refers to using multiple computers. They could be yours, rented, local, remote, clustured, desktops, servers, using any kind of framework. Cloud computing refers generally to renting remote servers from a third party who provide a particular software stack and interface and a restricted set of services. Yes the edges blur, but having 5 linodes to store data on is not really cloud computing, it's just renting servers, but having an S3 account is.
I think that was his point. Even though JavaScript had been used in web apps for years, as soon as "AJAX" caught on people (who didn't know better) started referring to anything that used JavaScript as "AJAX". I still get people asking me if such and such website is written in "AJAX".
Amen. For my little start-up in London to be able to unleash the computing power available on AWS has been amazing. We essentially buy desktop machines for the engineers and outsource everything else.
All our critical business apps are on Google and we can test and scale our software on AWS.
It's funny because the last company I founded (Electric Cloud) was specifically doing distributed computing stuff and we had to buy and install tons of servers just to be able to test our code. Today, founding that company would be totally different... we'd probably buy a few servers for some in house testing of key performance critical elements and then give our credit card to Amazon and buy what we needed as we needed it.
The combination of (for example) cloud + Hadoop is a giant slayer. So maybe he feels threatened. But the irony is that EMC, Larry's recent purchase, is making money hand over fist from cloud as it ramps up.
I don't really agree about Hadoop. MapReduce is interesting for some applications but it's not the only thing out there. We did extensive analysis of MapReduce at my new company and it clearly wasn't suitable for our needs.
However, the idea of moving code near to the data is a good one since moving data tends to be the slow part. But that's hardly a new concept, just look at Teradata.
I don't really agree about Hadoop. MapReduce is interesting for some applications but it's not the only thing out there. We did extensive analysis of MapReduce at my new company and it clearly wasn't suitable for our needs.
You say that as if MapReduce is the only thing you can do with Hadoop. It has a key/value datastore too, and since recently a relational database:
Just an example of something which benefits from inexpensive computing scale. Pretty narrow applicability. But it seems like there are plenty of other problems which also benefit.
It seems to me that what Larry's saying is that the idea of software as an on-demand service is nothing new. What he's objecting to is the hype around the "cloud" buzzword - as if it were some new invention and not simply internet-based, on-demand computing.
I can feel for him. I often find buzzwords are used to describe things I've been doing for 10 years in a way that makes it seem like people who only just discovered that thing are "more current" than me.
It's frustrating, and it actually comforts me that this can happen even to the likes of Larry Ellison.
Mana is something geeks quibble over when their magic missile cast fails.
Manna is magical honey-bread that falls from the sky for true believers.
Which is a funny word to use here, because you just wrote that comment to contradict Ellison's accusation that cloud computing was only for "true believers".
To me, Cloud stands for 'a service of which you don't know how it works'.
People have been drawing the Internet as a cloud for many, many years. No one knows how his packet actually gets delivered and that's what makes it so powerful. Even this simple piece of text goes through an intricate combination of radio waves, copper and fiber criss-cross through the world. but... I don't have to worry about any of it! All I have is this dead simple IP interface to my NIC (or URL interface to my browser) and I don't have to worry about whatever kind of nasty stuff goes on behind it. That is the most powerful idea in the history of information technology.
Cloud Computing is simply applying the principle to computing infrastructure. I don't know how EC2 works, I don't know where they get their machines, I don't know where they get their power, I don't know where they put my instances, I don't know who's managing the cabling, ... and I'm a happier man because of it.
When the term web 2.0 came out I hated it. It was just a description for what I was already doing. Similarly I didn't like the term cloud computing at first.
However the thing I have missed is that these terms are useful to explain to users what I am doing. If this is the case then these terms have value.
If you watch the video (which is pretty funny, and rather intelligent), his exasperation comes from the popular insistence that "cloud computing" is a new paradigm, and that businesses that don't use "the cloud" will die.
To somewhat paraphrase him: "All a cloud is, is computers and network (in terms of technology). In terms of business model, you could say that its rental."
Most of computing industry will remain unchanged, except maybe Dell's small business market, some of Microsoft's licensing revenue, and lots of desktop software that would be more productive as a network service.
Hype aside, there's no denying that the number of business offering and consuming services built on the cloud model have increased dramatically recently.
But no, instead of debating what the new services, platforms and business models are going to actually mean to the industry, let's quibble about the terms people are hyping for it.
When there's free publicity in giving old tech a new label, then of course companies are going to take advantage of that. Suppose you're a CEO and some interviewer asks you if you will move your software to the cloud, or interface with the cloud you pretty much have to advocate you have big plans for the cloud (even when it's nothing substantial) or you'll be put on the defensive. Since you can't persuade interviewers or the general public that cloud computing is meaningless for your business (if you already do SaS), it's a no win situation.
That said, I do think that cloud computing is innovation, in the sense that you can no longer point to an individual server and say "my software runs on that machine". We're now in an environment where -nobody- knows where your software is, and yet it still works.
Cloud computing is way overhyped and suitable for relatively few situations, but there is some innovation here.
If your main business consists of selling very expensive licenses on a per-node or per-cpu basis of a relational database then it makes perfect sense to speak these words. That doesn't make them true though.
The cloud is a real game changer, even though we are still working out exactly what it will do for us and how to deal with lots of issues (privacy, responsibility, reliability) it's got the established software businesses that came to power in the 80's and 90's quaking in their boots.
This will change the landscape of computing more than anything since '95, but it isn't going to happen overnight.
For every CTO that holds off a little longer based on Ellisons words the cash registers may ring one more time.
Key-value storage engines combined with cloud facilities go right at the heart of Oracles core business. What else is he supposed to say ?
Key-value storage engines combined with cloud facilities go right at the heart of Oracles core business.
Really? It seems to me that key-value systems are more of a problem for MySQL databases than Oracle ones. Pop quiz: how many people implementing key-value storage ever seriously considered buying an Oracle database?
I'm guessing very few.
I think what Ellison is saying is that cloud computing is nothing new. It's just a new name for distributed computing and we've had that for years. In fact, if you go back far enough we had SaaS on time-sharing systems: http://en.wikipedia.org/wiki/Time-sharing
They haven't considered buying an Oracle database when they're small, but when they grow now they don't have to either.
And bigger players are looking at switching from their expensive-software-on-big-iron to the cloud as well, and as a consequence will evaluate key-value storage systems vs rdbms, and some of them will switch.
What I'm arguing is that the trade off is key-value vs. MySQL for small companies. They pick their path at that point. If you go with RDBMS then you might end up at Oracle at some point. If you go with key-value then I guess you stay there.
If you've gone with key-value it's because you've decided that you don't need RDBMS features.
I'm just wondering about those cases where lots of data is stored in RDBMS which could have been stored in key-value storage engines as well. And with the increased query capabilities of k-v engines the no-mans-land is getting thinner.
The incentive to switch is equal to the license fee - the cost of the switch.
And possibly the convenience of being able to use cheap non-specialist hardware.
For the majority of the companies out there the cloud is still more expensive (and more risky, and privacy sensitive) than storing it on their own infrastructure.
But there are solutions in that space as well, having your own on-site 'private cloud', using all of the infrastructural tricks of cloud computing without the loss of control or the service premium.
For many big businesses I think that makes good sense.
Oracle has very reasonable license terms for virtualization. It's a gold mine for them.
When you license a host CPU, you can run as many oracle instances on as many VMs as you like. Oracle likes it that way: people spinning up "free" VMs for small projects drives demand.
"That magic part would connect the Network Computer to the Net. There would only be rudimentary software and memory on the Network Computer. Most software and serious memory would be out there on the Net where it could be easily maintained. The system would run on Java and use Oracle databases. Microsoft software would be nowhere in sight."
http://www.mondaymemo.net/031103feature.htm