Tech Policy Seminar

More On Bandwidth

I hope I'm not belaboring this point, but I've found evidence that contradicts the research I posted earlier this week claiming to show that there is little danger of Internet traffic exceeding capacity (and strengthens my original suspicion that such a "bandwidth crunch" is a likely limiting factor in the march toward Carr's World Wide Computer future).

Specifically, AT&T claims to be worried about increasing demand for bandwidth and investing in new capacity. It seems to doubt its own ability to keep pace, however.

According to the company's "VP for Legislative Affairs":

"The surge in online content is at the center of the most dramatic changes affecting the Internet today . . .  In three years' time, 20 typical households will generate more traffic than the entire Internet today."

And further that volume of internet traffic will increase 50 times by 2015, outstripping the ability of providers to keep up. As one might expect, video appears to be the real driver of demand.

Interesting stuff - but ultimately I don't know who to believe.

What Happens When We're Disconnected from the World Wide Computer?

Carr argues that we will in the future become so dependent on the "world wide computer" that we will not be able to live without it - that one disconnected will feel "lobotomized." That may not be a prediction, but an observation of current fact, at least for many people.

Social critics Parker and Stone have done some great work on this topic recently:

Update on Bandwidth Costs

I argued in a couple of posts last week that we should question Carr's assumption that bandwidth costs would stay low in the future. It looks like I was probably wrong. Here is an article that cites a lot of data supporting Carr's position. Basically, internet traffic growth is slowing while bandwidth growth is increasing. Note that these are second order measurements - traffic itself is still increasing year-to-year. This means that we are not relying on a finite amount of overcapacity (dark fiber) to absorb increased bandwidth requirements. To illustrate:

Data source: Telegeography. Image source:Ars Technica.

There is apparently still a bottleneck at the "last mile," which explains ISPs' concerns with P2P, but there apears to be no general shortage.

Processing power still might increase more rapidly (or decline in cost more rapidly) than bandwidth, but the central idea that bandwidth will get cheaper for the foreseeable future is probably right.

Hat tip to Slashdot.
 

Distributed vs. Utility Computing

(new longer version)

To elaborate on a point I made in my previous post, there seems to be a tension between the ideas in Carr's book about a "utility computing" future and a trend in the opposite direction, towards "distributed computing." Distributed computing is the idea that computing power will appear in more and more places, and make all kinds of devices more powerful or "smarter." Even more so than Carr's predictions, distributed computing is clearly happening already. Our cars are largely run by computers, as I pointed out earlier, my toaster has a microchip. Above all, one word - iPod.

The NY Times had an article yesterday on the trend, if you want to read more.

The point is this - is the distributed computing trend pushing against the forces Carr identifies as moving us toward utility computing, or can the two trends coexist? Coexistence is possible - imagine that some computing power is distributed to small devices in homes and businesses, putting it right where it needs to be used, while the remainder becomes provided over a distance by utilities. The big change from the current world is that the device between these two poles - the PC - disappears. Small jobs get performed at the point of action, and big jobs get outsourced. My guess is that this would be Carr's prediction.

I'm not sure it's right, however. Moore's Law makes computing power so cheap that even the chips in our toasters will be quite capable. It makes little sense to install a weak chip when, for essentially the same price, manufacturers can install a chip that outstrips a personal computer of only a few years ago. We may in the near future have a data center's worth of computing power in our homes, simply distributed across all of our devices (and the definition of device will expand - think smart clothes). With that much power in the home and in businesses, it will probably be much cheaper to harness it than to supply computing from a distance. The app that kills utility computing may be the guy that can get you ski jacket to talk to your microwave and work together to sort cells in a spreadsheet (or, of course, find aliens).

Essentially, this debate comes down to the relationship between bandwidth costs and processing costs. Carr's point depends on the economies of scale from centralization of processing exceeding the costs of distribution. My intuition is that processing is and will continue to be cheaper than transmission. I share Mr. James' skepticism about long-term decline in bandwidth costs, while we know for sure that the rate of increase in processing power per dollar is rapid and accelerating (Moore's Law). Carr's analogy to electricity (in my opinion, the best idea in the book so far, though it may not be original) is useful here - it makes sense to generate electricity centrally because the savings from doing so exceed the costs of transmitting it widely. This might not always be true - if cheap home solar power, fuel cells, cold fusion, etc. become feasible, it might be cheaper to generate power at home. A few people do this already, even contributing power back to the grid and being paid for it - though it doesn't make economic sense on a large scale. Just as cheap electricity generation technology would make electric utilities obsolete, cheap processing power might render IT utilities a stillborn idea.

A deeper analysis would go beyond these two inputs (generation and distribution costs) and look at other costs - storage and labor, in particular come to mind. Would economies of scale reduce these costs enough to counterbalance distribution costs, even if they do not reduce processing power costs? Maybe. It's an empirical question. It seems likely that labor costs would be a great savings - Mr. Bednarz' point in class that labor might be the greatest savings from centralization seems intuitively convincing. I would guess that storage tends to play a similar role to that of processing power - storage costs are decreasing at a similar rate - iPod storage size has increased even more rapidly than processing power.

In short, there are lots of inputs into the equation that determines the structure of the IT industry. They seem to cut in different directions, which both makes prediction difficult (leading me to question Carr's conviction that his model is right) and means that a variety of models are likely to exist. Change will occur, but it is possible for it to occur in many ways simultaneously - all prophets might be partly right and partly wrong. This makes me less worried about second-order issues like privacy (opt-out will be possible without being cut off from the world entirely) and natural monopoly (changing and coexisting models make dominance harder to achieve and less pervasive when it occurs).

Electricity ≠ The Interwebs

It's not my posting week (I guess I'd be really late if it were), but I just wanted to say a bit about Carr's analogy that the shift to remote computing (the term I'll substitute for "utility" computing) is somewhat like the move towards centralized electricity-generating facilities.  I'd also like to discuss the natural monopoly concerns with remote computing, as Carr himself suggests the possible need for regulation in a utility computing world on p. 61. 

The Electricity Disanalogy

At first the analogy to electricity might seem counter-intuitive simply because the internet has traditionally been considered a decentralized system, after all, that is supposedly it's genius.  But as Carr describes, there are significant benefits and efficiencies to centralized remote computing.  For a multitude of reasons though, I don't think this will lead to anything like the utility model of distribution we see in electric grids. 

There are three main differences (though they're all related) between what I see as the Electric Network and Carr's vision of "utility" computing. 

1) Electricity is a homogeneous product, remote computing/storage/software is not.   

2) The Electricity network is made up of essentially 2 components: generation and distribution.  Remote computing involves what I'd argue are 3 components: the underlying network (distribution), storage/computing capacity (hardware), and software. 

3) The Electric Grid carries exactly one homogeneous product: AC current.  The Internet, for the most part, similarly carries one homogeneous product: Packets.  However there is one huge distinction: packets can become whatever you want them to be.

The basic premise of the three points I'm trying to make is that there's a lot more diversity with what you can do over the internet network than what you can do with the electric network. 

Also, the electric grid might not be as centralized as it seems.  Carr described Google's new plant, which was located near a river to provide cooling to servers, thats a decentralized manner of getting power.  If you've ever been to California, you'll notice many people have solar cells on their roofs, and some technologies even enable you to share excess power (for those rare individuals to make more solar than they consumer) with the grid.  Can I write a book about how electric grids will one day mirror the internet's decentralized nature as wind and solar replace fossil fuels? 

The Lack of Natural Monopoly

If remote computing does explode, I'm not convinced that it will be that drastic a shift other than the fact that hard-drive and processor manufacturers will sell more equipment to hosting facilities rather than to direct consumers.  In particular I really don't think this will be a situation of natural monopoly. 

True, there are large fixed costs with developing a hosting facility, and there are economies of scale in such a facility, but there are a lot of areas for such facilities to compete, particularly when it comes to the software and services themselves.  Carr thinks IT is a commodity, if so, it's definitely one facing robust competition, especially when it comes to corporate IT services.

For one thing, we've been discussing privacy a lot, and when it comes to protecting another person's data there is a lot of room for competition. Redundancy and Latency are also areas for competition, how often is stuff backed up?  What are the maintenance practices?  These things are all true about the electric grid too, we all want reliable electricity as much as we want reliable remote computing, but with remote computing we actually have the ability to choose which remote computing providers we want to go with. In the electric grid, you go with your local power plant and local provider, in remote computing, you have your network provider, and THEN you can choose which remote computing service and host to go with.  [perhaps network neutrality is even more important in a world of remote computing, lest the network operators choose the winners of the remote computing market]

Consider Web-Site hosting services, they're very much like remote computing services, except that they play a more passive role.  They've existed for many years, involve high fixed costs in establishing the necessary server space and network connections, yet there are hundreds to choose from.  I think remote computing would most closely mirror this market.

 

I tried not to ramble... don't know if I succeeded.

    

Utility computing and democratic participation

A few of the posters (Somers and James) have noted concerns about privacy with the move towards utility / cloud computing.  Specifically, large data centers will contain detailed information about consumers that will allow companies to market products to consumers with surprising precision.  I believe Wal-Mart already has terabytes of data collected on consumer shopping habits.

While these are certainly fair concerns, I argue that utility computing could lead to too much privacy for consumers.  The privacy I discuss is that of seclusion and solitude, which can be furthered by technological connectivity.  This seclusion is antithetical to our democratic notions of civic participation.  I also note that pervasive utility computing exacerbates net neutrality concerns, which may also have anti-democratic effects.  Combined with the likely outcome of a single, dominant firm in a natural monopoly situation, consumers may be hampered in their abilities to pursue free speech, interaction, and civic participation.  Finally, I argue that regulation may not be as effective in this arena.

Social Interaction and Democratic Participation

Utility computing may lead to“privacy” of viewpoints and of interactions.  Carr notes that cheap electricity, cars, and gas prompted an exodus of the American middle-class from cities to the suburbs.  (P.97) These Americans forfeited public entertainment through parks and theaters and opted instead for private entertainment through radio and television.  We see that trend continuing today with the growing use of the Internet.

Various commentators have begun to express concerns about the effects of technologies, from television to video games to the Internet, on general human interaction and more importantly civic participation.    The rush to suburbia began to seclude families from vibrant cities full of people engaging each other to living in single-family homes with little interaction.  The Internet has done so on a more micro scale.  Individuals now rarely interact with others, except maybe through an occasional email or bulletin board post.  Professor Picker even noted that he had considered conducting our seminar completely over the Internet.  One can go days without speaking to another human being thanks to the slew of technologies in our lives that have cut out the human middle man.  In Japan, for example, there has been an alarming increase in the number of teenagers who stay in their rooms for years—or even a decade—without ever coming out.  They are fearful of social interaction.

While there is certainly concern about the minimization of face to face contact in modern society, the more troubling aspect is little human interaction at all and the possible pernicious effects that may have on a democratic society.  The theory of republican government as in the US system relies on civic virtue—the willingness of citizens to subordinate their private interests to the general good.  (Stone, Con Law casebook, p.12) Dialogue and discussion among the citizenry were critical features in the governmental process.  (Id.)  The model for government was the town meeting, with such political participation (in the form of discussions and not just voting) inculcating virtue that would allow subordination of the private interest to the public good.  (Id.)  The expansion of the Internet and thus solitary lives runs counter to this entire political foundation for a democratic nation.

The move to utility computing could easily worsen this facet of modern life.  Workers would not have to go to work.  With the availability of large amounts of bandwidth and terminals at home, there would be no need to.  Presumably, such unification would not just be related to work life.  At home, all things could be connected to each other and connected to some central computing unit.  A connected refrigerator would monitor items and automatically order relevant foods based on information stored at a central database about the user’s food purchases.  One would never need to leave the house or interact with another!  It’s not to say that most people have intellectual conversations about the state of the nation at the water cooler or when they visit their local grocery store, but the path to solitude is not always an obvious one.  Small amounts of removal may lead to a general distancing from society.  It would be a changing of social norms.  Just as increased availability of home appliances actually increased expectations about cleanliness in the home, increased levels of distancing from other humans in the smallest aspects of life would change expectations about general human interaction and discourse.

One could argue that the Internet and this general connectedness that Carr advocates would increase the amount of interaction and discourse.  The Internet allows people to meet and converse cheaply.  One can email, post on forums and blogs, etc.  Undoubtedly, the Internet may increase such interactions.  In my comment last week, I noted that the Internet had become a democratic forum for speech but probably could use some rules and limits, just as we have in the tactile world.  Without these limits, the type of interaction that would result on the Internet is unlikely to be beneficial for civic participation.  Sunstein, in Republic 2.0, noted the polarization effects of the Internet.  With so much data and information available at one’s fingertips, people generally only gather that data that fits their viewpoints and discard any opposing views, thus facilitating a move towards the extreme ends of a spectrum about some viewpoint.  With a two-way human interaction, each person is forced to confront the arguments of the other.  In a one-way interaction with technology, this simply does not occur.  Sure these large data centers could store all our information. But combined with the possible change in social norms toward solitude noted above, such information, for example, could only serve the one-way function of determining the ads we see.

Additionally, Strahilevitz has noted how the level of discourse plummets with the anonymization of speech on the Internet.  Those in the legal field saw these effects with websites such as AutoAdmit.com, where anonymous Internet speakers made vitriolic statements with little substantive value.  While utility computing does not require anonymous speech, it certainly makes it more possible technologically.  Anonymous speech certainly has its value, as with the Federalist Papers, but it also has its place.  That place may not be the whole of the connected realm.

Net Neutrality

There is one additional democratic concern relating to a connected world of utility computing: net neutrality.  Some worry that an Internet provider could limit the free flow of ideas on the Internet by controlling what content is delivered to the end user—either by slowing it down or charging more for access to it.  With utility computing, when many aspects of one’s life will be connected to a central server controlled by some single corporation, such worries are heightened.  If this company can control access to many different types of information depending upon its personal views, the ability of an end user to adequately participate democratically may be severely curtailed.  For example, if all our data is stored on central servers, what would prevent the company controlling this data to decide to limit access to political videos made by one’s friends?  Before, even if Comcast limited access through the Internet, the friend could still burn the video on a CD and share it.  With utility computing, the long term goal seems to be only to have “terminals” with all storage and processing happening on the backend.  That seems like quite a bit of control to hand over.

Natural Monopoly

One could argue that in a competitive marketplace, if consumers demanded net neutrality, companies would be unable to limit access in such pernicious ways.  The problem here is that utility computing, as others have mentioned, could plausibly lead to a natural monopoly.  Mr. Richardson and Mr. James have noted some of these monopoly concerns by pointing out that utility computing, just like electrical utilities, relies on large up-front fixed cost investments in capital with low marginal costs.  This situation is ripe for a natural monopoly.  The major non-economic concern here is that a single dominant firm is less likely to respond to consumer desires since it faces no competition.  Furthermore, the dominant utility computing firm need not worry about consumers substituting away to similar products or no products at all, since the society Carr envisions would be one with pervasive computing that touches on every aspect of people’s lives.

I am unconvinced by Mr. James’s counterargument that the hardware and know-how for utility computing could be widely distributed.  If there are real economies of scale here, as Carr argues, then this dispersion among firms will be inefficient and eventually uncompetitive in the face of one large, major competitor.  Likely, the outcome will be that this major competitor will simply buy up these small owners of capital.  Mr. James’s second argument is also unconvincing.  Assuming he is correct that providing of utility computing simply requires combining small individual pieces rather than purchase of a few large pieces, a natural monopoly is still likely to result.  Because this is a situation with seemingly large economies of scale, those who are able to combine the highest number of these small pieces into the largest central computing system will beat out any small competitors.  And those are able to do so will be those with the most amount of capital.  Therefore, only one firm is likely to survive. 

While the small pieces could spur entry, it is unlikely to because entrants know they will be unable to compete since they do not have the appropriate economies of scale.  With any threat of entry, the large, dominant firm could simply set a price below the costs of the entrant firm, since it has a large competitive advantage due to economies of scale.  There would be some concern about predatory pricing but that is unlikely to play out since the dominant firm could still be pricing above its own costs, which are so low due to economies of scale.  (See, for example, LePage’s.)  Finally, even if there were to be entry, it would lead to an overly competitive market similar to what happened in the railroad industry.  Firms would have undertaken large fixed cost capital expenditures but those would be seen as sunk costs.  Therefore, with competitiveness in the marketplace, they would price close to marginal cost and be unable to recoup any fixed costs, possibly leading to bankruptcy.

Regulation

The net neutrality and natural monopoly concerns point in one direction: regulation.  As Mr. Richardson notes, there are some standard regulatory tools such as rate regulation and price caps for dealing with natural monopolies.  I think, however, such regulatory fixes will be more difficult in the province of utility computing.  The appropriate rate would need to be set by determining costs of the monopoly and level of demand for its product.  Because of the pervasiveness of utility computing, the latter may be difficult to determine.  Demand will be influenced by so many different factors in the economy that I imagine it would be near impossible for an agency to be able to gauge.  Similar problems existed with electricity rate regulation, which was highly inefficient.  When secondary markets came about for the exchange of electricity, there was a large push for deregulation.  Rate regulation may also hamper adoption of utility computing in innovative ways and in new arenas.

It may also be difficult to determine costs because they are likely to be changing so frequently, as is often the case with computer hardware / software.  The bigger concern here, though, is that government regulation would hamper innovation that can lower costs for firms.  After all, if the rate is set to guarantee a rate of return given certain costs, where is the incentive to find new ways to lower them?

Conclusion

            Utility computing may be a great—and inevitable—march of progress in modern society, but there are several concerns that need to be tackled.  I have noted the specific one here about possibly changing social norms and consequently harming our democratic institutions.

Ahead on all fronts

The wonder of Nicholas Carr’s book, the Big Switch, is not in its ability to predict future changes of computing, but in making the reader realize how far towards completing those changes we really are. These changes are not making the front pages, but there are signs of them everywhere.  Yesterday, in the back pages of the Wall Street Journal, a small headline announced Google’s release of App Engine, a tool that allows software developers to run and build their web applications on Google’s technology infrastructure. One of the many benefits of the new service is that it saves companies and developers the headache of rearchitecting their websites and applications in order to accommodate increasing traffic. If successful in capturing the 10,000 small to medium size users that can use the beta version for free (down the road, Google plans make the service available to everyone and charge a small fee), larger companies should fall in line. Now that even one developer, via App Engine, can derive the benefits of Google’s technological infrastructure, companies with a technology infrastructures inferior to Google could find themselves at a competitive disadvantage.

Closer than we think

Google’s announcement and other similar services are evidence that Carr’s depicted transformation is happening even faster than we thought.  The service and others like it can capture users much faster than electric utilities could at the turn of the 20^th century. Whereas Samuel Insull had to convince the biggest users (industrial businesses) before he could go after the rest of the market , Google’s AppEngine, Amazon’s EC2 , and Salesforce’s AppExchange can capture any user, regardless of size, and charge accordingly (i.e., by the amount of space or traffic a particular website or application receives).

As more and more applications are created and used online, there will be less of a need for installing applications on a hard drive. The software model of generating revenue per number of installations/users (think MS Office) seems to be in serious danger, though companies with large resources should not have trouble moving from a download/buy and install model into an online access and use model. The trend towards online applications use and remote data storage facilities substantiates the terminal PC model Carr talks about in Chapter 4.

Consumers and manufactures fall in line

Developers and big companies are not the only players moving towards the utility model. Consumers are also falling in line. Apple’s new MacBook Air, the world’s thinnest laptop computer, is a good example. The strong buzz and demand for the laptop demonstrate that consumers are willing to adopt the “thin client” model, where a PC becomes merely a monitor that connects to the internet. The MacBook Air has no CD/DVD drive, no security slots, and no Ethernet ports. It really is more like a terminal that wirelessly accesses the internet.

The MacBook Air cannot be used alone; it functions primarily as a secondary computer. People who buy or intend to buy the device store or back up their files on a different device. This creates tremendous inefficiencies.  Even most without a MacBook Air use more than one computer and must transfer and/or update files on multiple computers. These days even files or photos on cell phones need to be synchronized. The problem compounds with the increasing popularity of Macs and the difficulty of transferring files between PCs and Macs. For example, someone who owns a Mac will usually have to use PC at work or at school.

These inefficiencies has been addressed by companies like SugarSync, which provides a web-based service that synchronizes and updates all your devices with the latest changes, whether they’re Macs, PCs, or Treo phones.  Even our law school has Webshare, also a web-based file sharing service, to address the multiple computer problem and save students the hassle of having to put a document on  a USB drive or e-mailing it to themselves in order to print from the Law School printers (not much of an efficiency gain since a user must down load the file in order to make changes and re-upload it again once the file is changed. Same hassle, different format). Nevertheless, as security, service, reliability and bandwidth increases, it will be difficult to resist the efficiency and cost savings of keeping, accessing, and backing up all files and applications online rather than storing them, installing them, and synchronizing them on multiple devices. As others have pointed out, with the advance of technology and the concomitant drop of data storage costs, we may reach a point where it may not be cheaper to have data stored remotely than on our own computer. But the advantage of utility computing is not just in costs – its foundation also lies in the convenience/efficiency of being able to plug into the grid from anywhere at any time while retaining access to the necessary data and applications one may require or desire. 

Computer manufacturers are already adapting. They have committed to developing a new category of computers that fit Carr’s terminal model. Netbooks, as they are called, are small portable devices with small screens, no hard drives, and minimal storage capacity. They run on Linux rather than Microsoft’s more expensive and memory intensive Windows. Netbooks allow user to perform minimal tasks, but they are essentially terminals that allow people to access the internet wirelessly. Ten computer manufactures have committed to making 20 different models with projected sales of 50 million annually by 2011 (see Netbook link). The terminal model of computing appears to be arriving even before the move towards utility computing/web-based services is complete.

Mutual reinforcement is speeding up the switch

The most surprising aspect of Carr’s book is that the trends he depicts are being advanced in all fronts: data/utility companies are continuing to increase capacity as well as web-based application/development platforms; developers and small to medium size companies are using the capacity and platforms of companies like Amazon and Google in addition to subscribing to software-as-a-service businesses like Salesforce.com and Netsuite; and consumers are playing their part by buying the platforms conducive to utility computing such as the Macbook Air and Netbooks. The three movements are mutually reinforcing; if they can converge effectively, it has hard to see how the move towards utility computing and web-base services can fail. Whether we like the big switch or not, Carr may be right when he says, "personal choice has little to do with it" (23).

 

If you build it, they will come, but how much will they pay for a ticket?

I had dinner this summer with a friend of mine who works at VMWare. He tried to explain to me the idea of virtualization. “Wasn’t that the big scam Enron tried to pull in the late nineties?” I replied, nonplussed. He responded that at an abstract level his company enables something similar for PC hardware. I, in turn, appreciated his willingness to talk at a level of generality that would not make me appear to be a complete idiot.

Indeed, the difference between the two examples—once you get beyond my friend’s generous generality—is telling. The Enron example supports Carr’s notion that we have radically changed our way of thinking about bandwidth. It adds to the idea of the many switches that Carr addresses—the primary change he discusses being the move away from the client-server model to utility-supplied computing. Rather than call bandwidth a commodity, Carr makes the case that the revolution is in treating bandwidth as a utility. He relies strongly upon an electricity analogy throughout his book. Whereas “commodity” implies a sense of proprietary ownership—that the bandwidth will shift as a matter of trading and borrowing—“utility” imagines conglomeration and not just trading. General purpose technologies have broad application, the breadth of which comes about through capturing their potential economies of scale.

Just as the revolution in computing has changed our own labor makeup—increasing the demand for service-based occupations, Carr traces a parallel development in the IT world itself. He notes a shift from hardware to a services-based model. The ability to overcome capacity limitations (Grove’s Law, p. 58) does seem to unleash growth potential.

Some posters have expressed skepticism that the public would be open to such a shift. At one point, Carr makes an allusion that seems to capture some of this doubt. He describes the Dalles Google facility as looking like a nuclear power plant with its four cooling towers. We can imagine that the public might harbor beliefs about the dangers of centralized processing just as it does about nuclear power. Part of the problem with accepting new technology is, as one poster terms it “technophobia.” This idea also has parallels to electricity. As Carr points out in his description of the Chicago Fair, electricity struck people as “at once glamorous and dangerous, it seemed, in itself and even more so in its applications, like a mysterious, invisible force that had leapt into the world out of the pages of science fiction.”

At the same time, Carr and other posters have tracked the step-wise conditioning of the public to embrace new technologies. Carr has tracked the development of small-scale versions of bandwidth as utility. He has shown that in this realm the public is willing to set aside its squeamishness about privacy and its risk aversion to trying novel things. The popularity of Napster, the existing use of services-based technologies, and the widespread embrace of ecommerce has shown that the public eagerly adopts technologies that save it time, energy, and money with little concern over privacy or system-meltdowns. Perhaps the Y2K anti-climax has laid to rest many of our technophobic anxieties about catastrophe.

Just as the public already seems able to wrap its mind around such centralized IT systems, it also will likely not prefer this generation’s version of Burden’s wheel. Shifting to virtualization thus is not analogous to moving from cars to subways. Carr uses the analogy of an apartment building, writing, “With the use of virtualization in the utility model, that building can be divided into apartments that can be rented out to dozens of tenants. Each tenant can do whatever it likes inside the walls of its own apartment, but all of them share the building’s physical infrastructure—and all of them enjoy the resulting savings.” The analogy to the Internet as superhighway, therefore becomes more apt, rather than less. Each person continues to travel in her separate vehicle while enjoying the benefits of a smooth 6-lane road of a type that she could never build on her own. The beauty is that now, she doesn’t have to own an actual vehicle, but can ride in a virtual car that still provides the same experience but doesn’t require a garage or gas. It also has 1000x the horsepower.

This all comes from the economies of scale. Many posters have noted that economies of scale favor, promote, and reinforce natural monopolies, but have not touched upon the competing trend that Carr argues will result. One effect that Carr traces is the idea of democratization. His use of the amazon.com example shows the piggy-backing potential for small companies. He notes that “Amazon’s utility levels the playing field.” One can imagine that in the same way that sites like etsy.com and ebay have opened up the door to even the smallest sellers, treating bandwidth as a utility will also provide access to efficiencies that small companies can exploit. Perhaps we will even see the re-emergence of markets that were not able to survive under the old model—etsy has shown us that the Internet can reinvigorate the market for handcrafts. This technological innovation might be surprisingly retro and unexpectedly egalitarian.

Can we imagine this all as a part of creative destruction? Can we imagine that the large companies providing the resources and economies of scale will feed the smaller entrepreneurs who will eventually destroy the old titans of past? Do the natural monopolies allow for a Schumpeterian kind of economic growth? Whether or not these promises, like the “Home of a Hundred Comforts,” will actually materialize remains to be seen. Our existing experience with the Internet seems to suggest that they are, at least, possible. 

Not to add too much to Carr’s already abundant use of analogies, but I am also reminded of Grameen Bank. Grameen, which means village, capitalizes upon economies of scale to provide microfinance. The small loans that they give and their methods of lowering overhead costs (largely through requiring social networks and debtor participation in enforcing debts) allows them to take on the kind of debtors that large institutions will not touch. They have democratized lending by reducing the costs associated with commercial banking. In some ways we can see utility-style communications technology as analogous to banking and credit. Pooling and centralizing assets allows a bank to make growth-generating loans. The more that we can manage costs (e.g. overhead and risk in the banking context) the more we can include smaller, "riskier" debtors in the credit pool. If we think of one obstacle to market entry as access to capital, increasing access (to capital, and by analogy, to IT power) can promote healthy competition and decrease the tendency to monopoly.

Yet Grameen also requires careful management to maintain its peculiar brand of commitment to microfinance—i.e. that which targets the best kind of sustainable growth for its small debtors. For example, part of its mission is to have a majority of its debt-holders be women. It also has rules about being involved, which require joining Grameen as part of a group. It also has rules about interest rates and what happens in the event of default.

In this vein, the Enron example might still give us pause about the inevitable commoditization of this utility. Carr skimps over the problems that we will inevitably face, hastily rattling off a list of challenges (p. 61).  In half a paragraph he gestures to the issues of metering, price-setting, load-balancing, diversity factors, different countries’ regulatory regimes, security, reliability, and efficiency. Frank, I think rightly, acknowledges that security, reliability, and efficiency issues have largely been resolved, and likely will continue to be more successfully addressed via the increasingly global competition between talented programmers. Frank also raises a telling point distinguishing issues that innovation will solve from thorny problems like the intrusion of governments. We might note that the technological innovation of the Insulls of the IT world may face a limiting factor more difficult to surmount with cleverness than Grove’s Law. They must navigate government regulation, which is even slower to change than network capacity. 

How do we make normative choices about the government’s role in managing this utility, whether it’s price-setting, regulation, and even load-balancing? Carr gives us no answers on how or whether to regulate this utility in the same way we’ve managed telephones and electricity. One might imagine that this particular big switch will require us to change our attitudes toward regulation. And as we have seen in cases involving microfinance, one might also imagine that our choices in how we regulate can affect who has access to the benefits of this technological innovation and how it affects competition generally.

(Insert pun on book title here)

Carr makes sweeping conclusions and writes with conviction. This makes for a good read, but also opens him to criticisms. Below are some significant ones, most at least touched on by other posters.

Distribution Costs:
As Mr. James notes, bandwidth may not be as cheap as Carr suggests, and may remain an obstacle to centralization of IT. Centralizing only makes sense if the savings from economies of scale exceed distribution costs. If the pipes are narrow, this is less likely to be true. This is essentially an empirical question, but it is certainly less clear-cut than Carr seems to think. It seems quite likely that distribution - not generation - will be the bottleneck for some time, if not forever. I don't think this is the case for electricity, and the difference strikes at the heart of the analogy on which the book is based. Still, I don't have the facts. Carr may be right here.

Privacy:
Mr. James and Mr. Bednarz note that people and firms might be less willing to use outside utility IT resources if they believe their secret data might become public in the process. This concern is larger in some contexts than others (the NSA and Swiss banks will likely never use utility computing), but seems real enough. As Carr points out, there were more or less rational fears about electricity as it was becoming a utility as well. In fact, an argument can be made that centralized computing might be more secure. Companies today have plenty of problems with data security, often due to a lack of good security practices. Centralized IT firms are likely to take security very seriously, and be quite professional about it (the sucess of their business depends on it). Certainly the damage caused by a breach at a big data center would be more severe than your average lost laptop, but if such breaches are much less frequent, the long-term average damage would be less. Think of it like driving vs. flying - plane crashes will probably kill lots of people at once, but driving is far more dangerous because accidents are more frequent. In short, I don't think privacy will be much of an obstacle to the trends Carr claims to foresee.

Costs of Commoditization:
All electricity consumers want the same thing. That might not, or might not yet be true with IT. I agree with Mr. Bednarz, though, that we are pretty close to that point already. Even to the extent that we are not, cheap computing power makes a great deal of customization possible. Mr. Cottrell points out that crossplatform code is much more inefficient - that is certainly true, but who cares when CPU power is so cheap? People give up their CPU time FOR FREE to FIND ALIENS! I run virtualization software on my Mac laptop to emulate MS-DOS. It takes a substantial slice of my system resources to do this, which seems silly when you consider how much more powerful my system is than a 386-class PC, but I don't care. Computing power, more than ever, is just about the cheapest resource on the planet. Providing different IT "goods" to different customers is therefore not a significant challenge.

Limited Advantages:
If computing power is so cheap, it ultimately limits the economy of scale advantages of centralization. (Mr. Cottrell touches on this point). For most applications, a single modern PC is more than enough to get the job done. Microprocessors now inhabit almost every consumer device, and are in almost all cases far more powerful than necessary.  My toaster has a microchip in it (the toast is still burnt, however). Sure, if you're Google, it makes sense to centralize the resources you use, and companies have long considered thin client setups, but why pay distribution costs (bandwidth, see above), when putting computing power where it is needed is so cheap? Consider this analogy: why don't we power our cars through the grid (presumably as plug-in electrics). The answer is simply that the internal combustion engine, while far less efficient than the steam turbines at power plants, is just fine when gas is cheap (as it remains, considering the energy contained within). In fact, the only reason plug-in cars are considered a good idea now is that we are beginning to consider the negative externalities of carbon emissions. If you pretend those don't exist - as we have been for the past century - centralization of energy for cars doesn't make sense. Just like cheap gas makes electric cars a poor option, cheap computing power makes central supply of IT much less attractive. Carr talks about Moore's Law, but doesn't give sufficient attention to how much it can undercut his predictions.

Regulation:
If Carr is right, it is possible that a single computing utility firm will emerge, with the power to restrict output and harm consumers. Certainly the tiny marginal costs (probably more important than the large fixed costs Mr. James mentions) of centralized IT generation make this possible. I am more optimistic than Mr. James is, however, about the ability of law and government to deal with the problem. Mr. James, it seems, would have had us keep the pre-Insull model of electricity generation because of the dangers of natural monopoly raised by utilities. If computing power is centralized as Mr. Carr predicts, and if a single firm emerges, the market will have clearly shown that the benefits of centralization are large. The costs to consumers can be kept down by the standard regulatory toolkit - rate regulation, price caps, divestiture etc. Sure, these can be inefficient, complex to implement, or even co-opted by industry, but would we really prefer the alternative?

Furthermore, as Mr. James also notes, the heterogeneity of demand and low entry costs associated with the market for computing make the emergence of a stable monopolist less likely than it would at first appear. In short, I'm not worried about this problem. It may not arise, and if it does, it is one we know how to manage.

So what?

Given these objections, where do I stand on Carr's predictions? I have to reserve final judgment until I finish the book. So far, however, I think Carr is right that centralized "generation" of IT is here and will be a big deal in the near future. This is noncontroversial, however, and is easy to see just by looking at Google Apps, Salesforce.com, and the long history of the thin client movement. The more radical version of his thesis, that there will simply be one World Wide Computer, seems much more questionable, given the more damning objections mentioned above. Still, I'll give Carr a fair shot to defend it in the second half of the book.

ComEd or Dan Ryan?

Nicholas Carr’s story is about the transformation of technologies—like computing and electricity—from something marvelous and awe-inspiring to ubiquitous, invisible features of everyday life. He also presses the ironic corollary: a technology is not truly revolutionary until no one sees as revolutionary any more. Electricity didn’t change society when it was worshipped in the White City, but when no one gave it a second thought. Market forces, in Carr’s view, make this transformation inevitable—economies of scale favor the centralized distribution of computing services, just like power, water, sewer and natural gas—and competition will thus force companies to adopt the cheaper utility model of computing, just as consumers will favor cheaper and more hassle-free utility computing for their personal use.  I want to expand here on issues of data security raised in other posts to question Carr’s analogy between IT and electricity and to ask whether Carr’s market-driven “technological imperative” will necessarily lead to widespread adoption of computing as a utility or whether other factors will lead to a less radical outcome.

Carr describes information technology and electricity as “general purpose technologies” (GPTs), which “are best thought of not as discrete tools but as platforms on which many different tools, or applications, can be constructed.” (15) People use data processing and storage to accomplish a wide range of tasks, just as electricity can be used to power many different tools. The question, though, is to what extent important differences are masked by this shared GPT label.

In the first place, electrical utilities send something of value to the user, for which the user pays, but computing utilities while similarly providing a valuable service are also something like data bailees. As Carr and several previous posts have noted, the computing utility company is entrusted with both trivial and extremely valuable and sensitive information. This two-way flow of valuable assets and the entrusting relationship between consumer and utility is an important distinction between utility computing and other utilities.

Second, the computing-electricity analogy may be obscuring something valuable as it seeks to displace the venerable internet-as-highway metaphor. A highway is a kind of general purpose technology, a platform adaptable to many vehicles, starting points and destinations; however, unlike a subway or light rail, users provide their own vehicles. At the risk of opening a topic beyond the scope of this post, let me just posit the conventional wisdom that many people around the world prefer driving their own cars, even when arguably more cost-efficient alternatives exist. See, for example, the Chinese appetite for cars, despite increasingly horrific traffic and air pollution in major cities. It is often said that individual ownership of cars provides a sense (though not always the reality) of freedom, autonomy, self-reliance, flexibility, privacy and safety—as compared to reliance on shared or public transportation alternatives.

I would argue that distributed (non-utility) computing provides many of these same attributes. A stand-alone computer offers the assurance of physical possession of data and of processing capacity. Of course, it seems probable that a computing utility will be able to use its economies of scale to develop superior hacker protection and more reliable redundant power sources and data backup. Perhaps data will in fact be more secure in such a facility. But many people who know that public transit is statistically safer still feel more secure while driving their own vehicles. Also, as others have observed, concentrated data facilities are a richer target for hacking, government surveillance and terrorist attacks on physical infrastructure.  The point is that the “technological imperative” does not seem, in the realm of personal transportation, to have necessarily resulted in the adoption of the least cost alternative. It is at least possible that the same will prove true for IT.

Carr opens his book with a Le Carrésque description of visiting VeriCenter’s Boston facility—an anonymous building in a run-down area, no sign on the door, a deserted entry with black handset, a heavy steel door; and then the high-tech wonders inside. I thought the cloak-and-dagger clichés made for a corny, if effective, set-up, and it occurred to me that just because an asset is extravagantly protected doesn’t necessarily mean it is valuable. But on further reflection, I think the anecdote drives home an important point: people are willing to pay a big premium to have their data secure. As noted above, computing utilities will have the scale to provide better security in some respects: hire the best IT people, invest more in hacker and virus protection, as well as power supply and data backup infrastructure. Their scale will give them an advantage over individuals and separate corporate IT departments. On the downside is the attractiveness of a computing utility to hacking, surveillance and physical attack.  Scale provides security advantages as well as downsides. I don’t have the expertise to assess which way the security balance will tip, or to guess whether, for irrational reasons, individuals and corporations will still prefer to drive their own computing vehicles. 

The other interesting aspect of the VeriCenter security story was the way in which the prosaic security measures of a spy novel protected an extraordinary facility. It represents a more complicated version of Carr’s story of technology. New technology may start as marvelous and utopian, and by becoming ordinary produce extraordinary social and economic changes, but at the same time, the new may coexist with, and be dependent upon, the old.  I think security concerns will most likely stymie a revolution in utility computing, producing instead a hybrid of utility and distributed computing, a mixture of extraordinary and prosaic solutions. This blending of new and old technology to ensure data security is illustrated by the following (prosaic) anecdote: a partner at my summer law firm advised me that notwithstanding metadata scrubbers, just to be certain, he still prints documents before sending them to opposing counsel and either uses fax or scans the printed pages and emails pdfs. Security concerns forced a hybrid of new and old technology, and this is my best guess as to how the contest between utility and distributed computing will play out.