January 28, 2008
In this interview with GRIDtoday, Nicholas Carr, author of the controversial book "The Big Switch: Rewiring the World, from Edison to Google," predicts what enterprise IT will look like as computing continues to move into "the cloud." He offers insights on the "revolutionary" nature of today's virtualization technologies, the future revenue model for IT vendors, and whether the Web will be able to handle the load.
GRIDtoday: To begin, can you briefly describe what will comprise “the big switch” and discuss how our interaction with computing will be altered in the years to come?
NICHOLAS CARR: By “the big switch” I mean a transformation in the nature of computing that’s going on right now, where we’re going from running our software and our computers privately and locally to more of a utility model where the computer functions we rely on -- whether at home or at work -- will be supplied over the Internet, which will become the equivalent of the electric grid for computing functions.
I think we can see this happening fairly quickly, particularly if you look at how individuals use their PCs these days; in essence, they’ve already made the big switch to online computing. Smaller companies are the next to make the move, and larger companies will probably be the slowest -- although in certain areas, such as software as a service, they’re making the move, as well.
Gt: How will this switch affect corporate computing and enterprise datacenters? Will we actually see big business adopt utility computing on a widespread scale?
CARR: I think for big businesses, the change probably will play out over a couple of decades. It’s not as if one day we wake up and everything’s changed and what used to be done locally is now done over the Net. It’s going to be a process of transition, where even very large companies are going to pull in more and more of the functions they need off the Web while also continuing to run their own datacenters and to run their own applications locally. I think we can see how some of the transformation will play out: companies will probably begin to bring in particular general business applications as services (the way they’re doing with Salesforce.com or Employease with HR applications), and at the same time they’ll be rebuilding their own internal IT operations to run more the way external utilities run. Over time, their own datacenters will merge with the general computing grid and, eventually, I think we can see a time when you’ll be able to run pretty much any software you need -- whether it’s a general software-as-a-service application or a customized application that you’ve developed yourself -- off the grid from utility datacenters. At that point -- and that might be a decade or two away -- all of computing begins to look much more like a utility.
Gt: Utility services exist right now, but companies aren’t exactly flocking to them. What will drive the increased adoption you foresee?
CARR: I think on one level it is cost. If you look at big companies ... current studies show that if they’re not flocking to the utility model, they’re certainly beginning to move there at a fairly rapid pace. If you move toward software as a service, not only are you replacing the capital cost of buying the software and licensing the software ... but you also avoid the need to buy the computers that run the applications, hire the people who maintain the computers that run the applications, and pay for the electricity and real estate necessary to house those computers. So it really represents, in many cases, substantial cost savings.
If you look at the big trends in big-company IT right now, you see this move toward a much more consolidated, networked, virtualized infrastructure; a fairly rapid shift of compressing the number of datacenters you run, the number of computers you run. Ultimately ... if you can virtualize your own IT infrastructure and make it much more efficient by consolidating it, at some point it becomes natural to start to think about how you can gain even more advantages and more cost savings by beginning to consolidate across companies rather than just within companies.
Gt: Do you see any differences between how companies are adopting, or will adopt, software as a service versus computing as a service? It seems as if software as a service is a bit more palatable right now.
CARR: A lot of it depends on the size of the company. If you look at small companies, they are moving fairly quickly toward computing as a service or storage as a service because they just don’t have the capital to build out their own infrastructures, so it’s very attractive. I think [with] the larger companies, it’s a matter of scale. A lot of big companies have enormous amounts of IT scale already, so until the utility suppliers can begin to match or exceed that scale, or begin to provide really significant savings, you won’t see the big switch to the grid computing model. (Editor’s note: Here, Carr’s mention of “a grid computing model” is in reference to an electricity-type utility model, not “grid computing” in the usual sense of GRIDtoday.) On the other hand, I think for a lot of companies that have spiky demand for certain applications or for certain computing jobs, particularly really high-powered computing jobs, it becomes much more attractive and much more natural to go out onto the grid and use that kind of capacity rather than building up their own capacity [for those applications or jobs].
Gt: Increasingly, I’m seeing software-as-a-service providers using computing-as-a-service models (e.g., Sun’s Network.com or Amazon’s EC2) to power their offerings. Is it just natural that a utility would utilize a utility in that manner?
CARR: It makes sense. It’s one way that the utility model bootstraps itself. If you’re starting off as a software-as-a-service supplier, being able to, in essence, rent a very sophisticated, very cutting-edge infrastructure, such as you can do through Amazon, is a much more efficient way to get your business going than having to raise and dump in enough capital to build up your own infrastructure and then scale it up if your company becomes successful. I think it’s natural for software-as-a-service companies to look to utility computing or utility storage as, essentially, their infrastructures -- and I think that will continue.
Gt: You mentioned companies virtualizing their own infrastructures. What are your thoughts on the current batch of service-oriented, virtualized models (e.g., grid computing, cloud computing, application virtualization, fabrics, etc.) that essentially allow companies to run internal utilities?
CARR: I think they’re revolutionary, really. It comes down to the basic reality of computing, which is that all digital computers are universal computing machines and you can translate hardware, which is in essence just software instructions that have taken physical form, back into software as long as you have enough computer power to run that software. I think that as computing power has exploded, it becomes much more feasible and much more economical to turn hardware into software. And not only that, but it becomes much more flexible and much less labor-intensive because you can do what you do with software all the time, which is automate things -- in this case, what used to be the physical hardware and application stack.
I think we’re just at the beginning of the virtualization wave, but, ultimately, it’s really going to transform the way companies think about their IT infrastructures internally, and, eventually, it’s all going to merge onto the grid.
Gt: Is this focus on less complexity, on-demand access to resources, dynamic provisioning, etc., essentially a middle ground between traditional siloed computing and the switch to external utilities?
CARR: I think it’s a step on the path more than just a middle ground. For companies ... it’s a way to fundamentally re-engineer their IT infrastructure to make it much more efficient. Once you’re able to virtualize the hardware on which you run applications, all of a sudden it doesn’t make any difference where the hardware that is running the virtualized machines exists. Once you’ve pulled all the costs out of virtualizing your own infrastructure, the next natural step is to move toward more of a pure utility infrastructure that is shared with many different companies.
Gt: So it’s pretty much a matter of easing in?
CARR: Yeah, and at the same time the external suppliers are gaining more scale, more maturity and more reliability.
Gt: Many people made the electric grid comparison with traditional grid computing, but not everyone bought the analogy. What makes your use of the electric analogy more accurate?
CARR: It’s not so much that it’s a different analogy, it’s just that technology advances, computing power gets cheaper and cheaper, virtualization technologies advance and the network gets more robust. It’s more a matter of timing: People have been talking about utility computing since the ‘60s, but it’s only recently that we really have the combination of cheap bandwidth, cheap computing capacity and virtualization technologies necessary to do it. Just as the ASP model in that late ‘90s didn’t particularly go anywhere, but now we’re seeing essentially a very similar model -- software as a service -- begin to take off, I think the same thing goes for computing grids. It’s just a matter of waiting until technology advances to the point where it becomes practical.
Gt: If, or when, “the big switch” becomes reality, what will it mean for the enterprise IT workforce?
CARR: If you look far enough ahead, and no one knows precisely when (Sun says that within 10 years it will be able to close down all of its datacenters and just run everything out of the cloud, and maybe that’s true or maybe it will take a bit longer), ultimately you don’t need your own datacenters, and a lot of the money and the people that companies now deploy for IT are no longer necessary. On the other hand, a lot of the traditional IT work -- whether it’s custom application development or thinking strategically about information management or how to automate and design processes -- all of those [might] become, if anything, more important simply because computing continues to get cheaper. Will those skills still reside in something called the “IT department?” I have no idea. It’s not a given that the IT department is something that has to last forever, and it certainly will change dramatically in terms of its role as this shift plays out.
Gt: What do you see happening to IT vendors, specifically those making a lot of money selling hardware or servicing their solutions?
CARR: We’re already seeing a lot of IT vendors begin to move fairly quickly to the new utility model, whether it’s Microsoft, SAP, Oracle, IBM, HP or Sun. I think it will be similar to what happened a hundred years ago when companies went from building their own generating stations to plugging into the grid -- some of the old vendors will make the transition successfully, some will be killed off, and some new companies will emerge.
The big question for the traditional existing vendors isn’t so much “Can we make the leap to this new model?” because for most of them, technologically, they certainly can. The big question is “Can we continue to make as much money with the utility model as we made selling the individual components to zillions of different companies running them all privately?” Right now, the indications are that it will be pretty tough to replicate the same levels of profitability and revenue that we’ve seen up to now in the IT business.
Gt: Speaking specifically about Sun, do you see this switch as a validation of what they’ve been preaching for years now?
CARR: For about 15 years! Sun’s problem is that they’re good at foreseeing the future, but they foresee it a little too early. I think Sun has been struggling in recent years with the commoditization of servers and other basic hardware, and it’s a high-end supplier so that puts it in a bind. But I think it sees the creation of these new, very efficient, very, very powerful centralized datacenters and utilities as a potentially very attractive market for its products and its expertise. Whether it succeeds or not remains to be seen, but it’s certainly a logical market for the company to focus on.
Gt: What kinds of companies do you see really being successful? Will we be accessing everything from Google and Amazon?
CARR: There are a couple of different models that could emerge, but any way you look at it there will probably be a fairly small number of companies that run the basic computing infrastructure -- for the very simple reason that we’re talking massive capital investments, the kind of multi-billion-dollar investments in the computing infrastructure that we’re seeing Google and Microsoft make today. It’s very hard, when you think about making money in that business, to think about a very large number of companies running these kinds of infrastructures.
The big question is “Will all the applications tend to consolidate among the companies that are running the infrastructure, or will we have a vibrant, competitive array of application suppliers that run on top of this utility infrastructure?” I don’t think we know that yet. What we know from [today’s] utilities is that if there is too much consolidation and centralization of power, then government tends to move in, and that’s certainly one scenario you could see playing out in this market, as well, if we see too much consolidation of control over the network.
Gt: Can the Web as it currently exists handle being used as the cloud not just for consumers, as it is today, but for enterprise users, as well?
CARR: Nobody really seems to know for sure, but I would say it will have to -- because too much is riding on it. We talk about Net neutrality a lot, and it’s a very important subject, but when push comes to shove, the Net more and more is the central infrastructure for business and commerce today; economically, we just can’t afford to have it fail. If that means cracking down on peer-to-peer networks and trading of big video files, then that’s going to happen to protect the commercial infrastructure, I can guarantee you, simply because we’ve come to a point where at an economic level -- and at a national security level -- you can’t afford to let the network fail.
Gt: How do you think the Internet will be able to handle, for example, the latency and performance issues that come along with certain enterprise computing jobs, such as those in the financial world?
CARR: If you look at the kind of responsiveness that we’re seeing from Salesforce.com, or even more so from Google (inside the Google search engine), you see that you already can achieve fairly high levels of responsiveness. That is only going to improve as software continues to be rewritten for these new parallel-processing or virtualized grids. I think it’s also important to remember that we’re not talking about total centralization of all data and all processing in distant plants. The fact is that we have in our local devices lots of computing power and lots of storage capacity, and I think software is going to be written to take advantage of that, as well as the centralized resources, through caching strategies and other kinds of software that make use of both local and centralized processing and storage capacity. You’ll have a very robust, vibrant system for running all sorts of software with responsiveness and speed that will be at least similar to -- if not even greater than -- running everything locally in a datacenter.
Gt: What steps are being taken, and what steps need to be taken, to get the Web ready to handle so much more information and so much more usage? Will the current fiber optic networks, for example, be enough?
CARR: Again, we don’t really know. We’ve seen in the last few years, particularly as the Web has turned into our central multimedia medium, an incredible growth in traffic -- and it continues apace and there continues to be large investments on the supply side in building out that infrastructure, and I think that will continue. An interesting question is “Will traditional telephone companies move into this area much more aggressively than they have in the past?” I think that’s a real possibility, as well. We certainly have an infrastructure that is in one way mysterious, since nobody really knows its capacity or its reliability, but, on the other hand, I think that as more and more commerce and activity move to it, we will continue to see large investments in it. Are they going to be able to move forward without any failures or any disruptions? That’s a good question, and I don’t know.
10/30/2013 | Cray, DDN, Mellanox, NetApp, ScaleMP, Supermicro, Xyratex | Creating data is easy… the challenge is getting it to the right place to make use of it. This paper discusses fresh solutions that can directly increase I/O efficiency, and the applications of these solutions to current, and new technology infrastructures.
10/01/2013 | IBM | A new trend is developing in the HPC space that is also affecting enterprise computing productivity with the arrival of “ultra-dense” hyper-scale servers.
Ken Claffey, SVP and General Manager at Xyratex, presents ClusterStor at the Vendor Showdown at ISC13 in Leipzig, Germany.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?