Upload
hoangbao
View
217
Download
4
Embed Size (px)
Citation preview
A special report on corporate ITOctober 25th 2008
Let it rise
CorpITCOV.indd 1CorpITCOV.indd 1 13/10/08 10:58:0113/10/08 10:58:01
The Economist October 25th 2008 A special report on corporate IT 1
Information technology is turning into a global �cloud� accessiblefrom anywhere, says Ludwig Siegele. What does that mean for the waypeople conduct business?
ing�became more distributed.Now, this special report will argue,
computing is taking on yet another newshape. It is becoming more centralisedagain as some of the activity moves intodata centres. But more importantly, it isturning into what has come to be called a�cloud�, or collections of clouds. Computing power will become more and more disembodied and will be consumed whereand when it is needed.
The rise of the cloud is more than justanother platform shift that gets geeks excited. It will undoubtedly transform the information technology (IT) industry, but itwill also profoundly change the way people work and companies operate. It will allow digital technology to penetrate everynook and cranny of the economy and ofsociety, creating some tricky political problems along the way.
Promise of heavenHere we go again, you may think. In orderto generate new demand, the maturing ITindustry keeps creating new buzzwords,often with celestial connotations (�cyberspace�, �blogosphere�), which suggestsome kind of technological nirvana. Thereality is much more down to earth.
Hype is indeed rampant in �cloud com
Let it rise
IN THE beginning computers were human. Then they took the shape of metal
boxes, �lling entire rooms before becoming ever smaller and more widespread.Now they are evaporating altogether andbecoming accessible from anywhere.
That is about as brief a history of computers as anyone can make it. The point isthat they are much more than devices in abox or in a data centre. Computing hasconstantly changed shape and location�mainly as a result of new technology, butoften also because of shifts in demand.
The �rst �computers� were indeed people. The word originally meant an individual who solved equations, often using amechanical calculator. Hundreds of themwere employed by big companies thatneeded to do a lot of numbercrunching,such as aeroplane manufacturers. It wasonly around 1945 that the word came to describe machinery.
But even after that, computing kept undergoing mutations�or, in the jargon, platform shifts. The mainframe, the originalcomputing platform, was dethroned byminicomputers, which in turn gave way topersonal computers, which are now beingpushed aside by handheld devices andsmartphones. With each step the architecture�the underlying structure of comput
An audio interview with the author is at
www.economist.com/audiovideo
A list of sources is at
www.economist.com/specialreports
Where the cloud meets thegroundData centres are quickly evolving into servicefactories. Page 3
Creating the cumulusSoftware will be transformed into a combination of services. Page 5
On the peripheryThe cloud’s communications with its clientswill become ever more intelligent andinteractive. Page 7
Highs and lowsAs IT gets cloudier, the economics of thebusiness will change. Page 9
The long nimbusThe cloud will make businesses moreadaptable, interconnected and specialised�and often smaller. Page 11
Computers without bordersThe cloud may be the ultimate form of globalisation. Page 13
Also in this section
AcknowledgmentsThe author would like to thank many people besides thosecited in the text for their time, ideas and kindness duringthe preparation of this report. They include MarcAndreessen of Ning, Brian Behlendorf of CollabNet, ChrisCapossella and Tim O’Brien of Microsoft, Simon Crosby ofCitrix, Rob Enderle of Enderle Group, Robert Hodges ofContinuent, Dirk Holzwarth of Alstom, GregPapadopoulos of Sun Microsystems, Dennis Quan of IBM,Simeon Simeonov of Polaris Venture Partners, Rahul Soodof SAP and Kishore Swaminathan of Accenture.
1
2 A special report on corporate IT The Economist October 25th 2008
2 puting�. The term entered into ITspeakonly a year ago and has spread voraciously. Cloud conferences and cloud blogsare multiplying almost as quickly as cloudstartups. Established IT �rms are slappingthe new label on old gear.
In fact, the cloud craze may havepeaked already, if the number of Googlesearches is any guide (see chart 1). Cloudcomputing is bound to go through a�trough of disillusionment�, as Gartner, aresearch �rm, calls the phase in the hypecycle when technologies fail to meet expectations and quickly cease to be fashionable. Much still needs to be invented forthe computing sky to become truly cloudy.
Yet even if the term is already passé, thecloud itself is here to stay and to grow. It follows naturally from the combination ofever cheaper and more powerful processors with ever faster and more ubiquitousnetworks. As a result, data centres are becoming factories for computing serviceson an industrial scale; software is increasingly being delivered as an online service;and wireless networks connect more andmore devices to such o�erings.
All this allows computing to be disaggregated into components�or �services�,in IT parlance. This is why European technologists such as Lutz Heuser, head of research at SAP, a German software giant,like to refer to it as the �internet of services�. The cloud metaphor seems moreapt. The internet is used mainly by peoplewith personal computers and a physicalnetwork connection. Cloud applications,on the other hand, will be used by billionsof devices of all kinds, many of them untethered, but will be connected to the �internet of things�.
In some ways the cloud is already hanging in the sky, especially for consumers. According to a recent study, 69% of Americans connected to the web use some kind
of �cloud service�, including webbasedemail or online data storage (see chart 2).The best example is Google, the biggest online search company by far, which now offers a plethora of webbased applicationssuch as wordprocessing or online spreadsheets.
Learning to �oatCompanies, too, have been moving intothe cloud, albeit much more cautiously. Financial institutions in particular have forsome time been building �computinggrids�. Firms that provide enterprise software as a service (SaaS) over the internet,such as Salesforce.com and NetSuite, havebeen growing steadily.
In the years to come companies are likely to venture much farther. For one, operators of computing clouds such as Amazonand Google have shown that this is a farmore e�cient way of running IT systems.Secondly, many �rms will �nd they haveno choice. The way in which their IT infrastructure has grown is proving unsustainable. Most corporate data centres today arecomplex warrens of underused hardwarethat require more and more people, spaceand power to keep them going. The currenteconomic malaise will increase the pressure on companies to become more e�cient. More has to be done with less, whichis cloud computing’s main promise.
This special report will chronicle therise of the cloud and try to predict where itis heading. It will start by looking at thetechnology. Computing clouds are immensely complex, but can be roughly divided into three layers: infrastructure, applications and the periphery where theymeet the real world. These will be discussed in turn. The report will go on to considerthe impact the cloud will have on the IT industry and the economy as a whole. Theconclusion will look at what might stopthe cloud from growing ever thicker: regu
lation and worries about the safety of bothpersonal and corporate data.
Irving WladawskyBerger, a technology visionary at IBM, compares cloudcomputing to the Cambrian explosionsome 500m years ago when the rate ofevolution speeded up, in part because thecell had been perfected and standardised,allowing evolution to build more complexorganisms. Similarly, argues Mr WladawskyBerger, the IT industry spent much ofits �rst few decades developing the basiccomponents of computing. Now thatthese are essentially standardised, biggerand more diverse systems can emerge.�For computing to reach a higher level�, hesays, �its cells had to be commoditised.� 7
2A young cloud
Source: Pew Internet &
American Life Project
*Survey based on 1,553 internet users†35% of internet users, connecting
to Wi-Fi away from home or work
Cloud-computing activities, by age cohortsApril-May 2008, % of users*
0 25 50 75 100
Web-mail services
Store personalphotos
Onlineapplications
Store personalvideos
Pay to storecomputerfiles online
Back up hard driveto an online site
Have done atleast one activity
Have done atleast two activities
All internet users/(All laptop users†)
56(64)
34(44)
29(38)
7(13)
5(10)
5(9)
69(79)
40(52)
Ages 18-29
30-49
50-64
65+
1Cumulonimbus
Source: Google Trends
Cloud computing, average weekly search volumesSeptember average 2007=100
S O N D
2007J F M A M J J A S
2008
0100
200
400
600
800
1,000
The Economist October 25th 2008 A special report on corporate IT 3
1
IT IS almost as easy as plugging in a laserprinter. Up to 2,500 servers�in essence,
soupedup personal computers�arecrammed into a 40foot (13metre) shippingcontainer. A truck places the container inside a bare steelandconcrete building.Workers quickly connect it to the electricgrid, the computer network and a watersupply for cooling. The necessary softwareis downloaded automatically. Within fourdays all the servers are ready to dish upvideos, send emails or crunch a �rm’s customer data.
This is Microsoft’s new data centre inNorthlake, a suburb of Chicago, one of theworld’s most modern, biggest and most expensive, covering 500,000 square feet(46,000 square metres) and costing$500m. One day it will hold 400,000 servers. The entire �rst �oor will be �lled with200 containers like this one. Michael Manos, the head of Microsoft’s data centres, isreally excited about these containers. Theysolve many of the problems that tend tocrop up when putting up huge data centres: how to package and transport serverscheaply, how to limit their appetite for energy and how to install them only whenthey are needed to avoid leaving expensive assets idle.
But containers are not the only innovation of which Mr Manos is proud. Microsoft’s data centres in Chicago and acrossthe world are equipped with software thattells him exactly how much power eachapplication consumes and how much carbon it emits. �We’re building a global information utility,� he says.
Engineers must have spoken with similar passion when the �rst moving assembly lines were installed in car factories almost a century ago, and Microsoft’s datacentre in Northlake, just like Henry Ford’s�rst large factory in Highland Park, Michigan, may one day be seen as a symbol of anew industrial era.
Before Ford revolutionised carmaking,automobiles were put together by teams ofhighly skilled craftsmen in custombuiltworkshops. Similarly, most corporate datacentres today house armies of �systemsadministrators�, the craftsmen of the information age. There are an estimated 7,000such data centres in America alone, most
of them oneo� designs that have grownover the years, re�ecting the history ofboth technology and the particular use towhich it is being put. It is no surprise thatthey are egregiously ine�cient. On average only 6% of server capacity is used, according to a study by McKinsey, a consultancy, and the Uptime Institute, athinktank. Nearly 30% are no longer in useat all, but no one has bothered to removethem. Often nobody knows which application is running on which server. A widely used method to �nd out is: �Let’s pull theplug and see who calls.�
Limited technology and misplaced incentives are to blame. Windows, the mostpervasive operating system used in datacentres, allows only one application to runon any one server because otherwise itmight crash. So IT departments just keptadding machines when new applicationswere needed, leading to a conditionknown as �server sprawl� (see chart 3).This made sense at the time: servers werecheap, and everrising electricity bills weregenerally charged to a company’s facilitiesbudget rather than to IT.
To understand the technology neededto industrialise data centres, it helps to lookat the history of electricity. It was only afterthe widespread deployment of the �rotaryconverter�, a device that transforms onekind of current into another, that di�erentpower plants and generators could be assembled into a universal grid. Similarly, atechnology called �virtualisation� now allows physically separate computer systems to act as one.
Virtually newThe origins of virtualisation go back to the1960s, when IBM developed the technology so that its customers could make betteruse of their mainframes. Yet it lingered inobscurity until VMware, now one of theworld’s biggest software �rms, applied it tothe commodity computers in today’s datacentres. It did that by developing a smallprogram called hypervisor, a sort of electronic tra�c cop that controls access to acomputer’s processor and memory. It allows servers to be split into several �virtualmachines�, each of which can run its ownoperating system and application.
�In a way, we’re cleaning up Microsoft’ssins,� says Paul Maritz, VMware’s boss anda Microsoft veteran, �and in doing so we’reseparating the computing workload fromthe hardware.� Once computers have become more or less disembodied, all sortsof possibilities open up. Virtual machinescan be �red up in minutes. They can bemoved around while running, perhaps toconcentrate them on one server to save energy. They can have an identical twinwhich takes over should the original fail.And they can be sold prepackaged as �virtual appliances�.
VMware and its competitors, whichnow include Microsoft, hope eventually toturn a data centre�or even several ofthem�into a single pool of computing,storage and networking resources that canbe allocated as needed. Such a �realtimeinfrastructure�, as Thomas Bittman ofGartner calls it, is still years o�. But the necessary software is starting to become available. In September, for instance, VMwarelaunched a new �virtual datacentre operating system�.
Perhaps surprisingly, it is Amazon, a bigonline retailer, that shows where things areheading. In 2006 it started o�ering a computing utility called Amazon Web Services(AWS). Anybody with a credit card canstart, say, a virtual machine on Amazon’svast computer system to run an application, such as a webbased service. Developers can quickly add extra machines
Where the cloud meets the ground
Data centres are quickly evolving into service factories
3Voracious
Source: IDC
Worldwide spending, $bn, on:
FORECAST
0
50
100
150
200
250
0
10
20
30
40
50
1996 98 2000 02 04 06 08 11
New servers
Server management and administration
Power and cooling Servers,installedbase, m
4 A special report on corporate IT The Economist October 25th 2008
2 when needed and shut them down if thereis no demand (which is why the utility iscalled Elastic Computing Cloud, or EC2).And the service is cheap: a virtual machine, for instance, starts at 10 cents per hour.
If Amazon has become a cloudcomputing pioneer, it is because it sees itself asa technology company. As it branched outinto more and more retail categories, it hadto develop a sophisticated computing platform which it is now o�ering as a servicefor a fee. �Of course this has nothing to dowith selling books,� says Adam Selipsky, incharge of product management at AWS,�but it has a lot to do with the same technology we are using to sell books.�
Yet Amazon is not the only big onlinecompany to o�er the use of industrialscale data centres. Google is said to be operating a global network of about threedozen data centres loaded with more than2m servers (although it will not con�rmthis). Microsoft is investing billions andadding up to 35,000 servers a month. Other internet giants, such as Yahoo!, are alsobusy building huge server farms.
In some places this has led to a veritabledatacentre construction boom. Half a dozen are being built in Quincy, a hamlet inthe middle of America’s Washington state,close to the Columbia River. The attractionis that its dams produce plenty of lowcostpower, which apart from IT gear is themain input for these computing farms. Onaverage, cooling takes as much power ascomputing. Microsoft’s new data centrenear Chicago, for instance, has three substations with a total capacity of 198MW, asmuch as a small aluminium smelter.
But cheap electricity is only one, albeitimportant, criterion for choosing the site ofa data centre. Microsoft currently feeds 35sets of data into an electronic map of theworld, including internet connectivity, theavailability of IT workers, even the airquality (dry air makes a good coolant), tosee where conditions are favourable andwhich places should be avoided. Apparently Siberia comes out well.
Google, for its part, seems to be thinking of moving o�shore. In August it applied for a patent for waterbased data centres. �Computing centres are located on aship or ships, anchored in a water bodyfrom which energy from natural motion ofthe water may be captured, and turnedinto electricity and/or pumping power forcooling pumps to carry heat away,� saysthe patent application.
Many chief information o�cers wouldlove to take their IT infrastructure out tosea and perhaps drown it there. Even as de
mand for corporate computing continuesto increase, IT budgets are being cut. At thesame time many �rms’ existing IT infrastructure is bursting at the seams. According to IDC, a marketresearch �rm, a quarter of corporate data centres in Americahave run out of space for more servers. Forothers cooling has become a big constraint.And often utilities cannot provide the extrapower needed for an expansion.
Fewer, bigger, betterSo IDC thinks that many data centres willbe consolidated and given a big makeover.The industry itself is taking the lead. For example, HewlettPackard (HP) used to have85 data centres with 19,000 IT workersworldwide, but expects to cut this down tosix facilities in America with just 8,000employees by the end of this year, reducing its IT budget from 4% to 2% of revenue.
Other large organisations are followingsuit. Using VMware’s software, BT, a telecoms �rm, has cut the number of servers in
its 57 data centres across the world from16,000 to 10,000 yet increased their workload. The US Marine Corps is reducing thenumber of its IT sites from 175 to about 100.Both organisations are also starting tobuild internal clouds so they can move applications around. Ever more �rms are expected to start building similar inhouse, or�private�, clouds. The current economicmalaise may speed up this trend as companies strive to become more e�cient.
But to what extent will companies outsource their computing to �public� clouds,such as Amazon’s? James Staten of Forrester Research, a marketresearch �rm, saysthe economics are compelling, particularlyfor smaller �rms. Cloud providers, he says,have more expertise in running data centres and bene�t from a larger infrastructure. Yet many �rms will not let companydata �oat around in a public cloud wherethey could end up in the wrong hands. Theconclusion of this report will consider thequestion of security in more detail.
It does not help that Amazon and Google recently made headlines with serviceinterruptions. Few cloud providers todayo�er any assurances on things like continuity of service or security (called �servicelevel agreements�, or SLAs) or take onliability to back them up.
As a result, says Mr Staten, cloud computing has not yet moved much beyondthe earlyadopter phase, meaning thatonly a few of the bigger companies are using it, and then only for projects that do notcritically a�ect their business. The Washington Post, for instance, used Amazon’sAWS to turn Hillary Clinton’s WhiteHouse schedule during her husband’stime in o�ce, with more than 17,000 pages,into a searchable database within 24hours. NASDAQ uses it to power its serviceproviding historical stockmarket information, called Market Replay.
Stefan van Overtveldt, the man incharge of transforming BT’s IT infrastructure, thinks that to attract more customers,service providers will have to o�er �virtualprivate clouds�, fenced o� within a publiccloud. BT plans to o�er these as a servicefor �rms that quickly need extra capacity.
So there will be not just one cloud but anumber of di�erent sorts: private ones andpublic ones, which themselves will divideinto generalpurpose and specialised ones.Cisco, a leading maker of networking gear,is already talking of an �intercloud�, a federation of all kinds of clouds, in the sameway that the internet is a network of networks. And all of those clouds will be fullof applications and services. 7
The Economist October 25th 2008 A special report on corporate IT 5
1
WANT to become a programmer in 20minutes? With the program from Ice
berg, a startup, you can. Just open the �getting started wizard� and pick the type ofapplication, say �project management�.Type in its �business objects��things like�client�, �team� and �members��and tellthe tool how they relate to each other.Then design some input forms and de�nethe process for getting a project done. Another click and you are ready to go.
In reality, of course, things are not thatsimple. And many professional programmers will sco� at the development tool asa mere toy. Yet Iceberg and similar out�tsdemonstrate that geeks are losing their monopoly on programming. Now, with a bitof patience, anybody can create a simpleapplication, for instance, to collaboratewith colleagues or to draw an online map.
This democratisation of programming,however, is only a small part of somethingmuch deeper: a fundamental change in thenature of software. It is not just that moreand more software will become a servicedelivered online. More importantly, applications, webbased or not, will no longercome as a big chunk of software, but willbe made up of a combination of electronicservices�a shift that has picked up a lot ofspeed since computing began moving intothe cloud.
To understand this new way of building applications, known as �serviceoriented architecture� (SOA), think of a culinary analogy. Whereas the old chunk ofsoftware resembles a precooked meal thatjust has to be popped into the oven, thenew architecture is more like a restaurant.It is a service in itself but also a combination of subservices. There is the waiter
who takes the order and conveys it to thekitchen. There is the cook who preparesthe food. And there are the cleaners whokeep the place tidy. Together they create the�application�: a restaurant.
An attack of the vapoursThe importance of this shift from a monolithic product to services is hard to overstate. In a sense, it has seeded the cloud, allowing the droplets�the services thatmake up the electronic vapour�to form. Itwill allow computing to expand in all directions and serve ever more users. Thenew architecture also helps the less technically minded to shape their own clouds,using such tools as Iceberg’s.
Just as for the industrialisation of datacentres, there is a historic precedent for thisshift in architecture: the invention of movable type in the 15th century. At the time,printing itself was not a new idea. But itwas Gutenberg and his collaborators whothought up the technologies needed tomake printing available on a mass scale,creating letters made of metal that could bequickly assembled and reused.
Similarly, the concept of modularityhas been around since the early days ofcomputing. �Everything in computer science is to just write less code. What is thetechnique for writing less code? It’s calledsubroutines,� said Bill Gates, Microsoft’sfounder, in a recent interview. A subroutine is a part of a program that can be reused, just like movable type. The idea, saysMr Gates, has always been to apply thisprinciple of a subroutine more broadly.
Yet this did not happen, mainly because the cost of computing fell much faster than that of communications. Ever
cheaper and more powerful chips made itpossible to move from mainframes tominicomputers to personal computers(PCs) and now to handheld devices. Butconnecting all these pieces remained di�cult and expensive, which meant that suchdevices all had to come with their owndata and chunky programs. Now, thanks toplenty of cheap bandwidth and more andmore wireless connectivity, computing isable to regroup into specialised services, orMr Gates’s subroutines: �We now live in aworld whereð[a] subroutine can exist onanother computer across the internet.�
Part of Gutenberg’s genius was to recognise the need for all the letters to be identical in height so they could be easily combined. Similarly, for computing services towork there had to be robust technical standards. Only a few years ago this seemed farbeyond the IT industry’s reach. Most �rmsinsisted on their proprietary technology,mostly to lock in their customers. Again,cheaper communications helped to bringabout change. The success of the internetdemonstrated the huge bene�ts of openstandards and forced vendors to agree oncommon ways for their wares to work together. One result is a stack of somethingcalled �webservices� standards.
Serviceoriented architecture �rstshowed up in opensource software butwas quickly adopted by big enterprisesoftware vendors because they had apressing need for it, says Jim Shepherd ofAMR Research, a consultancy. Big softwarevendors, for instance, had to �nd a way tountangle the hairball of code that their products had become, or else they themselveswould choke on it. Customers wantedmore �exible and extensible programs.
Creating the cumulus
Software will be transformed into a combination of services
6 A special report on corporate IT The Economist October 25th 2008
2 Think back to the gastronomic example. A precooked meal is hard to change,and so are traditional software applications. By contrast, a restaurant can easilychange its menu and its style of operation.Similarly, SOAbased software allowscompanies to alter their business processes, such as the way they handle orders tocollect cash.
SAP, a German software �rm, was oneof the �rst companies to put this serviceoriented architecture front and centre.Starting in 2003, it developed, among other things, a new corporatesoftware package that did away with monolithic applications, such as programs to keep track of acompany’s �nances or manage its relationship with customers. Instead, it introduceda collection of reusable components thatcould be strung together at will.
IBM, too, is a fan of SOA and web services. But its approach is somewhat di�erent, given that it does not sell business applications but makes most of its moneyfrom IT services and software to managethe underlying computing infrastructure.IBM uses SOA mainly to help �rms integrate their increasingly complex and disparate IT systems. Its software turns theminto a collection of services that can bewoven into business processes.
The approaches may be di�erent butthe vision is the same: to create IT systemsthat adapt to the business needs of companies and allow them to connect. �When Iwant to do something new�, explainsSteve Mills, the boss of IBM’s softwaregroup, �I do not need to build a new application but can use the pieces I alreadyhave.� To Peter Zencke, who led the development of SAP’s new package, its most exciting feature is that �any of the processcomponents can now become a serviceprovided by some other �rm.�
Despite millions of dollars spent onmarketing SOA, it has not really taken o�yet. But many webbased applications forconsumers rely on this concept. The primeexample is Google Maps. When the onlinegiant launched the service, programmersquickly �gured out how to mix the mapswith other sources of information. This ishow, for example, Housingmaps.com wascreated, a combination of a Google mapwith the rental and sales listings fromCraigslist, a website for classi�ed ads. Itwas one of the �rst �mashups�, as suchcombinations have come to be called.
Since then the number of such mashups has exploded, thanks mainly to services like Microsoft’s Pop�y and Yahoo!Pipes. In essence these are graphical pro
gramming tools. Users drag and drop�modules��data feeds providing such information as pictures, headlines andsearch results�and weave them together.
Most of these mashups are still toys,but �rms o�ering software as a servicehave started o�ering similar combinations. In April Salesforce.com and Googleannounced that they would integrate theironline services. Users of Salesforce, whichhelps �rms manage their customer relationships, can now quickly switch to Google’s webbased applications.
Smaller �rms have already started toweave a network of services. OpSource, aSilicon Valley startup, for instance, provides basic services for other SaaS �rmsand web companies. TriCipher, anotherCalifornian newcomer, authenticates users of web applications. Ribbit, for its part,allows these services to add voice communications to their o�erings.
Yet it is unlikely that the software cloudwill end up as a vast nebula of thousandsof specialised services. Even creating a serviceoriented architecture is �no silver bullet� against complexity, in a famous phraseby Frederick Brooks, an elder of computerscience. Although web services allow online o�erings to connect, for instance, it iscostly to synchronise their data. And itwould not make sense for any �rm to betits business on simple mashups.
As software markets mature, they tendto form two kinds of clumps: integratedsuites of applications, and platforms ontop of which others can build programs.Both forms are already emerging. On theapplications side there is Google Apps and
Zoho, which is even more comprehensive.It encompasses a total of 18 applications,including word processing, project management and customerrelationship management (CRM).
As for platforms, there are already plenty, in di�erent shapes and sizes. For enterprise applications, SAP has built one calledNetweaver. Oracle o�ers something similar called Fusion. Last year, Salesforcelaunched a �platform as a service�, allowing other �rms to use the plumbing thatsupports its own CRM o�ering.
More recently platforms for consumerservices have been proliferating. Facebook, a social network, was the �rst to become one in 2007. Other big online �rmshave followed suit or will do so soon: Google with App Engine, Yahoo! with Y!OSand Microsoft with a �cloud operating system� thought to be called Windows Strata.Some predict a platform war to rival theepic �ghts between Microsoft’s Windowsand Apple’s Macintosh.
Never say dieWhat shape will the software cloud take,other than being a vast collection of services? In one way it will look much like theold software world. There will be a few bigplatforms, akin to today’s operating systems, and most applications will be written to one of these platforms.
What is less clear is just how much ofbusiness and consumer software will migrate into the cloud, and how fast. The answer depends on whom you ask. Unsurprisingly, Marc Benio�, Salesforce’sfounder and chief executive, argues thatweb applications will spell the �death ofsoftware�. But people are not about tothrow out their powerful PCs or other �client� devices, if only because many ofthem still work o�ine at times. Similarly,companies will always want to keep someapplications inhouse, for reasons of security, regulation or simply to maintain control. Ray Ozzie, Microsoft’s chief softwarearchitect, promotes something called�software plus services�, meaning thatcustomers will settle on �the right mix ofold and new stu��.
If history is any guide, Mr Ozzie is moreon the mark. Even the biggest changes in IThave never spelt the death of anything,notes Josh Greenbaum of Enterprise Applications Consulting. IBM, for instance, isstill making money with mainframes.
So the software cloud, just like its hardware underpinnings, will be very diverse.But how will people make use of this kindof computing? 7
4Looming ever larger
Source: Gartner
*Within the enterprise application
markets †Estimate
Worldwide software-as-a-service revenue*, $bn
2005 06 07† 08 09 10 110
2
4
6
8
10
12
20.0
15.9
9.0
25.9
95.7
Compound annual growthrate, 2006-2011, %
29.1
Content, communications and collaborations
Enterprise-resource planning
Supply-chain management
Customer-relationship management
Office suites and digital-content creation
Others
F O R E C A S T
The Economist October 25th 2008 A special report on corporate IT 7
1
IT WILL take something with a lot morebang to replace a medium that is thou
sands of years old. That was the prevailingreaction when Amazon last November announced the launch of Kindle, an electronic book reader the size of a paperback thatcan store more than 200 volumes. Yet bythe end of this year Amazon will have soldnearly 380,000 Kindles, says Mark Mahaney, an analyst with Citigroup, a bank.�Turns out the Kindle is becoming the iPodof the book world,� he recently wrote in anote to clients, in a reference to Apple’siconic music player.
It is certainly not the Kindle’s looks thatexplain its success. Compared with theiPod, its design looks very last century.Software and battery life, too, leave a lot tobe desired. The chief attraction of the device is the ease with which it can be used tobuy books and other content. Equippedwith a mobilephone modem, the Kindlecan simply pull new reading material outof the air. Users do not even have to have awireless service contract. �Our vision is tohave every book that has ever been inprint available in less than 60 seconds,� explains Je� Bezos, Amazon’s boss.
It remains to be seen whether the Kindle will become a cultural phenomenonlike the iPod, of which around 160m havebeen sold so far. Amazon, for its part, isdownplaying the Kindle’s success and willnot con�rm any sales estimates. But it issafe to say that, once the next generation ofwireless networks is up and running, hundreds of millions of devices will come, likethe Kindle, with builtin radio connectivity(see chart 5). Digital cameras will automatically upload pictures. Smart meters willsend readings of how much electricity ahouse consumes. All kinds of sensors willbe able to send messages, even things likedipsticks when tanks of liquid are low.
The relationship of these devices tocloud computing may not be obvious. Butif huge data centres and applications makeup the cloud itself, then all the hardwareand software through which it connectsand communicates with the real world areits periphery. In IT speak, this is known asthe �front end� or �client side�.
As the Kindle and other examplesshow, this layer does not have much to do
with the user interface or client device ofold. It will do a lot of computing itself. Itwill come in all shapes and sizes, depending on what the user wants to do. And itwill not just distribute information, as theweb does, but collect it as well. The analogy that springs to mind here is a theatreperformance with audience participation:the electronic cloud will adapt to whateverit engulfs.
As you like itJust like computing itself, the dominantuser interface has evolved continually. Inthe days of the mainframe, when computers and their peripherals �lled entirerooms, people communicated with thesemachines �rst via punch cards and thenvia greenglowing monitors, which weresimply dumb terminals. Only with the riseof personal computers did the user interface become more intelligent, responsiveand graphical.
The �rst version of the web was thus abrief step backward. To be sure, browsersbrought colour and graphics to the hitherto textbased internet, but they were asdumb as the mainframe terminal. This haschanged only in recent years. A bundle ofwebdevelopment techniques dubbedAJAX and multimedia software such asAdobe’s Flash and Microsoft’s Silverlightnow allow programmers to write what arecalled �rich internet applications� (RIA).
Whatever the buzzword, the principleis much the same. Servers no longer dishup simple hypertext markup language(HTML), the web’s early lingua franca. Increasingly, web pages are bona �de piecesof software that are executed in the browser. Users of Web 2.0 sites who venture intomenu items such as �view source� in theirbrowsers can sometimes see thousands oflines of code.
In recent months the browser has become even more of a platform for otherprograms, akin to an operating systemsuch as Windows. The main driver of thistrend is Google, with its huge strength indistribution that can only gain from moreand more software being o�ered as a service. In May 2007 the Silicon Valley �rmlaunched Gears, a program that allowsweb applications to be used o�ine, and inSeptember this year it released a newbrowser called Chrome. Its most important feature is that it can execute several sophisticated web applications at once.
Although for now the internet browserwill remain the main vehicle for people tointeract with the cloud, other forms arecoming to the fore. One is the �widget�, asnippet of code that often lives on a PC’sdesktop and allows the user to get a quickpersonalised view of a set of data. The ideais that a salesperson, for instance, shouldnot have to �re up an entire application forcustomerrelationship management to�nd out which leads to follow up.
More importantly, there is now a greater variety of hardware through which toaccess the cloud. Already, desktop and laptop computers are starting to lose their monopoly for sur�ng the web as smaller devices such as smart mobile phones andvarious forms of portable computers startto compete with them.
Asus, a Taiwanese computermaker,started the trend when it launched a small,cheap laptop called �Eee� a year ago. Nowthere are dozens of these devices. Gartnerreckons that 5.2m of these �mininotebooks� will be sold this year, 8m next andas many as 50m in 2012.
Perhaps the best indicator of things tocome is Intel, a huge chipmaker. It made afortune selling processors for servers, personal computers and laptops. In June the
On the periphery
The cloud’s communications with its clients will become ever more intelligent and interactive
5Chatterboxes
Source: iSuppli
Worldwide shipments of electronic productswith built-in wireless capability, bn
2003 04 05 06 07 08 09 10 11 120
0.5
1.0
1.5
2.0
2.5
3.0
Local-area
Products with embedded wireless networking:
Wide-areaTotal cellularhandsets
F O R E C A S T
8 A special report on corporate IT The Economist October 25th 2008
2 �rm launched a new line of chips calledAtom, designed to power what it calls �netbooks� and �mobile internet devices�(MIDs), mainly intended for sur�ng theweb. Intel is also the driving force behindWiMAX, a technology for wireless broadband access to the internet. It wants to puta WiMAX radio chip into as many devicesas possible, from portable computers tospecialised gadgets such as the Kindle.
Apple’s iPhone and its App Store,which allows iPhone and iPod owners todownload applications, also provide aforetaste of how important wireless devices will be for the cloud. Apple launchedApp Store only in July. Two months later ithad already tallied 100m downloads,meaning that it took o� much faster thanApple’s highly successful iTunes musicstore. Many of the programs on o�er connect to the cloud, including news feeds,multiplayer games and a service thatkeeps track of the latest polls for America’spresidential election.
You can take it with youThe plethora of devices wirelessly connected to the internet will speed up a shiftthat is already under way: from a �devicecentric� to an �informationcentric� world,in the words of VMware’s Paul Maritz. Upin the cloud there will be a body of data foreach individual that will accompany themthrough life, he explains, and it will not betied to any particular device, as it is today.
Again, what will make this possible is
virtualisation�this time of client devices,not servers. With the help of software fromVMware and others, some �rms have already virtualised their employees’ desktopcomputers, which allows them to be managed centrally. Operating systems and applications will no longer run only on theemployee’s PC but on a virtual machine ina data centre that can be accessed remotely,theoretically from any PC in the world.Sooner or later mobile devices will also become virtualised. Users will be able to usetheir applications and data on whichevergadget they have at hand.
Yet the cloud’s interface is designed notmerely to provide information but to gather it as well. The future belongs to servicesthat respond in real time to informationprovided either by their users or by nonhuman sensors, predicts Tim O’Reilly, thefounder of O’Reilly Media, a publisher oftechnology books who coined the term�Web 2.0�. Such �live applications�, hesays, will get better the more data they areable to collect�and there will be plenty asthe cloud expands.
One of the �rst examples of such a service was Google. What originally put thesearch service ahead of the competitionwhen it was launched a decade ago was itsway of harvesting the information provided by web users in linking to other sites:the more links point to a page, the moreuseful it must be. These days most links aregenerated by computers, so the originalform of this �page rank� algorithm has
long since been scrapped. But Google’s approach is still the same: mining information provided by web users, such as theirsearch histories, to provide more relevantsearch results and more e�ective and targeted advertising.
The direct link to users also allows �rmssuch as Google continuously to improvetheir interface, something traditional softwaremakers were not able to do. At anygiven time Google is running dozens oftests to optimise the look and feel of its offerings. This makes web applications farless technologydriven and much moreuseroriented, says IBM’s Mr WladawskyBerger. �They are much more inspired bywhat goes on in the real world.�
A raft of startups is also trying to builda business by observing its users, in e�ectturning them into human sensors. One isWesabe (in which Mr O’Reilly has invested). At �rst sight it looks much like any personal�nance site that allows users to seetheir bank account and creditcard information in one place. But behind the scenesthe service is also sifting through its members’ anonymised data to �nd patterns andto o�er recommendations for future transactions based, for instance, on how much aparticular customer regularly spends in asupermarket. Wireless devices, too, will increasingly become sensors that feed intothe cloud and adapt to new information.
Nokia, for its part, is planning to buildall kinds of sensors into mobile phones tomonitor things like movement, barometricpressure or even the owner’s health,which many experts expect to become abig new trend. Sensors could also be usedto record people’s activities, creating whatsome already term a �lifelog��raising allkinds of privacy concerns.
As wireless technology gets better andcheaper, more and more di�erent kinds ofobjects will connect directly to the cloud.SAP, the German softwaremaker, haslaunched a research project called �The Internet of Things� to see what can be donewith the resulting information. As part ofthat project, an initiative called the �FutureFactory� is now under way to investigatehow intelligent tags can make manufacturing more adaptive and e�cient.
More and more data get you only so far,however. In the end, Google’s search results and its textbased online advertisements are relevant to users only becausethe �rm has devised clever ways to siftthrough them, says Hal Varian, the �rm’schief economist. The big challenge of thecloud will be to connect the myriad data init and make them pro�table. 7
The Economist October 25th 2008 A special report on corporate IT 9
1
EVEN elephants can die. In 1993 extinction came close for IBM, then the
world’s largest computermaker (it hassince been overtaken by HP). Its mainframe business was collapsing and pro�tswere plummeting. At that point LouisGerstner took over as chief executive andmanaged to turn the company around.�Only a handful of people understandhow precariously close IBM came to running out of cash in 1993,� he writes in hismemoir, �Who Says Elephants Can’tDance?� �Whether we would have had to�le for bankruptcy, I can’t say.�
There are many reasons why IBM nearly went bellyup, not least the fact that BigBlue had become a bureaucratic monster.But most critically, it had failed to adapt tothe industry’s �rst big platform shift,which only really made itself felt in the early 1990s: the move from mainframes tosmaller machines, �rst socalled minicomputers, then personal computers. �IBMwas slow, very slow, in delivering distributed computing, and many small companies moved in to �ll the gap,� Mr Gerstnerwrites.
It is unlikely that the move into thecloud will produce a similar neardeath experience or even a real casualty�if only because IBM still stands as a warning. Butthat does not mean that the structure of theIT industry will remain unchanged, northat the economics will stay the same.Once the IT sky really clouds over, individ
ual �rms’ share of the global IT budget (seechart 6) will shift.
The move to distributed computing,which started in the mid1980s, led to a bigchange in the IT industry. In the era of themainframe computing came in a verticallyintegrated package, mainly from IBM. Withdistributed computing the industry became a stack of horizontal layers. In corporate IT these were mainly hardware, thenetwork, infrastructure software (such asoperating systems and databases), enterprise applications and IT services.
Not all of these layers were createdequal. Computermakers commanded athicker one, for instance, but software companies were more pro�table. The key program was the operating system, both onservers and on personal computers (�clients�). It was the standard to which othercomponents of IT systems had to conform.Usually this was a version of Windows,which made Microsoft the IBM of this newera of computing.
Cloud computing is unlikely to bringabout quite such a dramatic shift. In essence, what it does is take the idea of distributed computing a step farther. Still, it willadd a couple of layers to the IT stack. One ismade up of the cloud providers, such asAmazon and Google. The other is softwarethat helps �rms to turn their IT infrastructure into their own cloud, known as a �virtual operating system for data centres�.
Drawing a neat diagram of the IT stack
will also become increasingly di�cult because the layers are becoming less distinct.In a world of services it often does notmake sense to think of hardware and software separately, argues Padmasree Warrior, the chief technology o�cer of Cisco.Both need to be blended to o�er new services, she says.
Mix and matchEven though the IT stack may not changeall that much, the perceived value of thedi�erent layers will shift, and with it theamount of pro�t IT �rms can make fromeach of them. Who will lose and who willwin depends on how much of computing
Highs and lows
As IT gets cloudier, the economics of the business will change
6Follow the money
Source: Gartner *Forecast
Worldwide end-user spending, $trn
2006 07 08* 09*0
1
2
3
4
Computer hardware
Software
IT services
Telecomsequipment
Telecoms services
10 A special report on corporate IT The Economist October 25th 2008
2
1
eventually moves into the cloud. In the �rst round almost everybody in
the IT industry will do well as the cloudsare being built. The biggest winners arelikely to be hardwaremakers, says MarkStahlman of Gartner: �Hardware alwayswins when new demand for computing isuncovered. And we haven’t had such asweeping global demand since the 1990s.�
But in the longer term there will be relative winners and losers. The hardwarebusiness could actually �nd itself in thelosing group. Its margins could getsqueezed as cloud computing matures because there will be fewer customers withmore buying power, says James Staten ofForrester Research. Large cloud providerscan dictate how to build servers and atwhat price, he notes.
All that may explain why hardwaremakers were among the �rst to jump onthe cloudcomputing bandwagon. So far,they have done only what Mr Staten calls�cloudwashing�: relabelling existing products that help customers build a more�exible IT infrastructure. But they are alsopreparing for a time when more moneycan be made building clouds than buildingcomputers. IBM and HP, for instance, haveteamed up with other �rms and universities to design new cloud architectures.
Which side of the fence?In the long run, says Mr Staten, hardwaremakers may be torn between supplyingcloud providers or becoming providersthemselves. Being both will not be easy,because the �rms would be competingwith their biggest customers. Dell seems tohave decided to be a cloud supplier. SunMicrosystems is a candidate to become aprovider; it is o�ering a cloudlike servicecalled Network.com, albeit not very successfully. HP and IBM, already used to thebalancing act of selling hardware and providing IT services, will try to do both.
Makers of traditional software will �ndthe going even tougher. With the advent ofopensource software, in particular Linux,selling operating systems had already become less pro�table. In a virtual worldthey will become even more commoditised, which is bad news for Microsoft.Many business applications no longerneed a big, generalpurpose server operating system but can use a specialised one,which should put pressure on prices. Onclient computers, more and more applications are written to run in browsers, not onany particular operating system.
Makers of business applications arealso on the defensive. Traditionally they
have made billions by selling their programs, often demanding hefty sums to install them and then charging an annualmaintenance fee for upgrades and technical support. But this highly lucrative business model has come under increasingpressure, says Michael Cusumano, a professor at the Massachusetts Institute ofTechnology (MIT).
For one, he says, software vendors willhave to �nd new ways to charge for theirwares: in the cloud, tying licensing fees tothe number of users, for instance, will bedi�cult, since services will mostly be consumed by other machines. More importantly, the corporate world has becomeless and less willing to buy software forlarge sums of money, so software �rms listed on America’s stockmarkets now makemost of their pro�ts from maintenanceand other services (see chart 7). SAP will increase its annual maintenance fees to atleast 22% of a program’s value over thenext few years, in line with those of Oracle,its main rival.
Yet the biggest challenge for software�rms is to become providers of online services themselves, says Brent Thill of Citi Investment Research. So far they havemoved slowly, o�ering SaaS only on theside, if at all. This was partly because theircustomers were not that keen. But moreimportantly, notes Mr Thill, the softwarehouses are still wedded to their old business model. With SaaS they do not get a bigupfront payment, only subscription fees.
Once Salesforce and NetSuite hadshown that the SaaS model works, the incumbents began to move faster. In September last year, for instance, SAP presented�Business ByDesign�, a package of webbased enterprise applications for smallerbusinesses. But success will not come easily. SAP has slowed down the introductionof the new service because it still needs towork out how to run it cheaply enough tomake a reasonable pro�t.
Pure SaaS providers also have a lot ontheir minds. Some experts, such as JoshuaGreenbaum of Enterprise ApplicationsConsulting, reckon that few will ever be aspro�table as traditional software �rms. Although it is almost a decade old, Salesforcestarted making money only in 2006, mainly because it �rst had to spend heavily onmarketing to attract customers. But nowthat the service has 1m users and revenuesof more than $1 billion, these costs willcome down, says the �rm.
The companies that have the bestchance of making money from the cloudare those that get things to connect and
work together and help customers movetheir computing around. This is music tothe ears of big IT �rms, not least IBM. Nearly 80% of its revenues come from infrastructure software and IT services, which itcan o�er globally. HP is catching up, havingtaken over EDS, another big ITservices�rm. Both Microsoft and SAP, for their part,believe that �rms will want to have achoice in where to do their computing, aswell as the �exibility to move thingsaround over time.
Two potentially important contestantsare rarely mentioned: Cisco and EMC, theleading makers of networking and storagegear respectively. Having invested a lot insoftware and services, Cisco has becomemore than just the source of most of theworld’s routers, the tra�c cops of the internet. It is betting that in the cloud the network layer will become more important,for instance to ensure that computingworkloads are able to move around securely. EMC, for its part, has made two dozen cloudrelated investments andlaunched a cloudinfrastructure division.
Whoever manages to own the dominant operating system for the data centrecould become a big winner. VMware isbound to have a shot at this. As well as being the market leader in virtualisation, ithas the support of EMC, which holds 86%of the �rm. But the competition is likely tobe intense.
Will this prospective platform war produce a dominant company in the mould ofIBM or Microsoft that is able to extractmore than its fair share of the pro�ts? Probably not, because it will be relatively easyto switch between vendors, says GeorgeGilbert of Tech Strategy Partners, a consultancy. Nor is it likely that one �rm willmanage to build a global cloud monopoly.Although there are important economiesof scale in building a network of data cen
7Serving up a treat
Source: Michael Cusumano,
Massachusetts Institute of Technology
US-listed software-firms’ sales, % of total
1990 92 94 96 98 02 042000 06
0.3
0.4
0.5
0.6
0.7
Services
Products
The Economist October 25th 2008 A special report on corporate IT 11
2
1
�BUSINESSES, as well as most organisations outside the business world,
begin to shift from hierarchical processesto networked ones. Nearly every facet ofhuman activity is transformed in someway by the emergent fabric of interconnection. This reorganisation leads to dramaticimprovements in e�ciency and productivity.� So said Wired magazine, the centralorgan of Web 1.0, in July 1997 in an essay entitled �The Long Boom�, arguing that theworld was in for �25 years of prosperity,freedom and a better environment�. Backthen, the article re�ected the general opti
mism that led up to the internet bubble.Now, after two busts, several wars andgrowing fears of global warming, it makesfor somewhat surreal reading.
The cloud lends itself to similar hyperbole. Yet so far there has not been much debate about its economic fallout�probablybecause the �new economy� ended badlyand the newest one is currently doing evenworse. There will be many ways in whichthe cloud will change businesses and theeconomy, most of them hard to predict, butone theme is already emerging. Businessesare becoming more like the technology it
self: more adaptable, more interwovenand more specialised. These developments may not be new, but cloud computing will speed them up.
Corporate IT has always promised tomake companies more agile. In the 1990smany companies reengineered their business processes when they started using aform of software called enterpriseresource planning (ERP), which does thingssuch as managing a �rm’s �nances andemployees. But once these massive software packages were in place, it was exceedingly di�cult to change them. Implement
The long nimbus
The cloud will make businesses more adaptable, interconnected and specialised�and often smaller
tres, the computing needs of companiesand consumers vary too widely for onesize to �t all.
Even if the cloud is likely to transformthe IT industry, some things will stay thesame. One is the importance of lockin. Ifanything, companies and developers willbe even more dependent on cloud platforms and applications than they are onthe old kind. SaaS promotes the �hollowing out� of IT: a �rm that needs to migrateto another system will no longer have therequired expertise. When Facebook, say,makes a change to its platform, developershave no choice but to go along with it.Some are already calling for a �CloudComputing Consortium�, in the mould ofthe World Wide Web Consortium (W3C),to set standards that allow applications tomigrate easily from one platform to another. One standard initiative, called �OpenSocial�, already allows the same webbased application to run in several socialnetworks, which are also clouds of sorts.
But standards go only so far. Some fearthat one company could try to monopoliseother key parts of the cloud; ironically, Microsoft worries that Google is doing exactly that with the online advertising market.To Steve Ballmer, Microsoft’s boss, Google’s advertising platform is like a �ywheelthat picks up speed as more websites attract more advertisers, and vice versa.
Eric Schmidt, Google’s chief executive,denies any evil intent to achieve worlddomination. He argues, with some justice,that it would be hard for Google to controlthe cloud, if only for technical reasons:much of it is already based on open stan
dards, and its loose structure does not lenditself to locking customers in.
Mr Schmidt promises that Google willnot lock its users in either. �Our competitive advantage is not from lockin�, he says,�but from having specialised knowledgeof how to build data centres and how tobuild new software that is not reproducible, such as our search algorithm. This ishow we make our money.�
Yet Google is more like Microsoft than itlikes to admit, says Nicholas Carr, a technology writer and blogger. Microsoft, heargues, achieved its dominant position inthe PC world not least by commoditising
products, such as the browser, that arecomplementary to its cash cows, such asWindows: as their cost came down, demand for Microsoft’s products went up.
Similarly, Google’s natural instinct is todo its utmost to encourage people to spendmore time online, because that will givethe company more opportunities to selladvertisements and collect data aboutthem. According to Mr Carr, almost everything the company does�building hugedata centres, �ghting copyright restrictions,digitising the world’s libraries, developinga new browser and, most recently, evenhelping to launch satellites�is aimed at increasing the use of the internet. �Googlewants information to be free�, he recentlywrote in his blog, �because as the cost of information falls it makes more money.�
But Google may never become as powerful as Microsoft because regulators areunlikely to let it. Microsoft was eventuallyput in the dock for abusing its monopolybecause it got too greedy, pushing most ofthe rest of the industry to complain. Giventhat the world has already lived throughthe Microsoft drama and that Google willa�ect many more industries, the searchcompany is likely to be restrained muchearlier. The �rm is currently in negotiationswith the US Justice Department about acontroversial advertising partnership between itself and one of its competitors, Yahoo!, which would further strengthenGoogle’s position in online advertising.
Even if the economics of the cloud arestill in �ux, though, it is already clear that itwill have farreaching implications forbusinesses and for society as a whole. 7
12 A special report on corporate IT The Economist October 25th 2008
2
1
ing SAP, the market leader in ERP, is likepouring concrete into your company, goesan old joke among IT types.
This helps to explain why in many�rms IT departments and business unitshave traditionally been at loggerheads. Inrecent years tensions have worsened.Companies must grapple with everchanging markets and regulations, yet IT budgetsare being cut. Many �rms now have a hugebacklog of IT projects.
Hence the interest in cloud computing.It turns capital expenditure into operational expenditure, which makes things mucheasier and cheaper. Instead of having toshell out a lot of money for, say, a server totest an application and, even with luck,wait a few weeks for it to be up and running, managers just have to whip out acredit card, open an account at AmazonWeb Services (AWS) and �re up a virtualmachine for a few dollars.
In many �rms senior managers probably do not even know that business unitsare using AWS or similar services. In someorganisations, however, it has already become a muchappreciated R&D tool. Pharmaceutical companies, for instance, areregularly tapping into AWS to calculatesimulations. Sogeti, a European consultancy, has used a cloud built by IBM to allow itto test new ideas and quickly put togetheran IT system for a companywide brainstorming event.
Cloud services have also been hugelysuccessful with startups, which can nowenjoy infrastructure of the same quality aslarge companies. In fact, AWS is probablythe main reason why there are now somany �rms o�ering all kinds of �Web 2.0�services. Their usefulness may sometimesbe hard to gauge, but that is a good thing. Itis a sign of lively �combinatorial innovation�, made possible because entrepreneurs can cheaply try new combinationsof technology, says Google’s Mr Varian.
Many startups would probably noteven exist without the cloud. Take Animoto, a service that lets users turn photosinto artsy music videos using arti�cial intelligence. When it launched on Facebook,a social network, demand was such that ithad to increase the number of its virtualmachines on AWS from 50 to 3,500 withinthree days. �You could give me unlimitedfunding,� says Adam Selipsky of AWS,�and I wouldn’t know how to deploy thatmany servers in 72 hours.�
Combinatorial innovation should alsobe made easier by the fact that the cloudwill be a huge collection of electronic services based on standards. But this service
oriented architecture will be even moreimportant for existing �rms because itshould free their inner workings�their�business processes��from the straitjacketof their ERP systems and allow these processes to be more easily adapted, for instance to launch a new product.
Again, the software industry has beenpromising this for some years under thebanner of serviceoriented architecture,discussed in an earlier chapter. Yet theadoption of SOA has been slow and manyprojects have failed, says Chris Howard ofthe Burton Group, a consultancy. The reasons are not just technical but cultural; forexample, some business units are not usedto sharing data. Cloud computing will helpresolve some of these problems. Manywebbased services are built to be integrated into existing business processes.
Don’t do it yourselfWhat e�ect will all this have on the natureof the �rm? If IT systems really allow companies to become more modular and �exible, this should foster further specialisation. It will become even easier tooutsource business processes, or at leastthose parts of them where �rms do not enjoy a competitive advantage. Companieswill increasingly focus on their �core� andshed the �context�, in the words of Geoffrey Moore, managing director of TCG Ad
visors, a consultancy.This also means that companies will
rely more on services provided by others.They will increasingly form �process networks�, a term for loosely connectedgroupings of specialised �rms coined byJohn Hagel, a business strategist at Deloitte& Touche, an auditing �rm. His prime example is Li & Fung, a company based inHong Kong that has assembled a globalprocess network of nearly 10,000 businesspartners in the clothing industry fromwhich it puts together customised supplychains for clothes designers.
Both trends could mean that in futurehuge clouds�which might be called �industry operating systems��will providebasic services for a particular sector, for instance �nance or logistics. On top of thesesystems will sit many specialised and interconnected �rms, just like applicationson a computing platform. Yet this is onlyhalf the story. The cloud changes not onlythe plumbing and structure of �rms andindustries, known as the �transactionallayer�, but also their �interactional layer�,a term coined by Andy Mulholland, chieftechnologist of Capgemini, a consultancy.He de�nes this as the environment whereall the interactions between people takeplace, both within an organisation andwith its business partners.
Despite all the technology that has en
The Economist October 25th 2008 A special report on corporate IT 13
2
1
READERS may be excused if they havenever heard of Kinakuta. It is a tiny is
land dreamed up by Neal Stephenson inhis novel �Cryptonomicon�. In the science�ction classic, libertarian technologists and entrepreneurs try to turn it into adata haven for such things as anonymousonline banking and electronic money.
Reality has indeed caught up with fantasy. Iceland is about to become a Kinakuta
of sorts. Data Islandia, a local company, istrying to establish the island as a vault for agrowing pile of data that �rms must retainin order to comply with all kinds of regulations. It has a compelling pitch. With itscool climate, abundant geothermal energyand secure remoteness, Iceland appears tobe a prime location for data archives.
As often, however, truth is strangerthan �ction. In a way, Data Islandia is erect
ing borders in the cloud: it intends to storeEuropean data according to European regulation and American bits according toAmerican rules. What is more, to keep thedata safe during transport, they are pickedup with a �data scooter� (in essence a container �lled with disk drives) and taken toIceland by aeroplane�as though �breoptic links had never been invented.
This illustrates the political tensions
Computers without borders
The cloud may be the ultimate form of globalisation
tered the workplace in recent years, so farthis layer has not really changed. PCs certainly made people more productive, butmost of their programs were not designedfor collaboration. The enterprise applications they worked with were still centralised systems. And email has in some waysmade things worse as the �ood of messages takes up lots of time and attention.
The dominant model is still that people�rst labour individually and then mergetheir respective e�orts, says Mr Mulholland. �It’s not much di�erent from the ageof paper,� he writes in his book �Mesh Collaboration�. �Collaboration often meanspulling up your chair next to your colleague so you can look at the same screen.�
Consumers have pulled ahead of companies in using cloudbased services thatallow for better collaboration, such asblogs, wikis and social networks. The �rstgeneration of people that has grown up using all these tools is now entering theworkforce. Being used to a culture of sharing information freely, these �digital natives� will be impatient with the rules oftraditional corporate IT.
So it is helpful that �rms have at last begun to embrace Web 2.0 technologies inearnest, a trend predictably called Enterprise 2.0. By 2013 companies around theglobe will spend $4.6 billion on such tools,according to Forrester Research. What nobody knows is how �rms will trade o� theadvantages of letting employees collaborate with the outside world against the associated risks, for instance that con�dential information is leaked. Because of suchsecurity concerns (see chart 8), many �rmsblock access to such sites as Facebook oncompany computers.
Companies may not have much choicebut to open up, says Mr Mulholland. Em
ployees will increasingly resist constraintson their use of technology, and they willhave a growing need to reach beyond thecorporate �rewall. Twenty years ago, he argues, 80% of the knowledge that workersrequired to do their jobs resided withintheir company. Now it is only 20% becausethe world is changing ever faster. �We needto be open to new and unknown connections with people and content,� he says.
Exploiting the �mesh�, in the words ofMr Mulholland, will also mean that employees may build simple applications oftheir own that allow them quickly to automate repeat tasks, using internal and external IT services. A new product by SerenaSoftware called Mashup Composer gives aforetaste of things to come. Using a visualinterface, even the not so technically minded can quickly put together a service todeal, say, with travel requests or approvedocuments.
Yet the impact of the cloud will also befelt on a macroeconomic level. Just as itmakes small �rms more competitive, itwill help developing economies to move
ahead. �The biggest promise is that it willmake computing cheaper and easier touse�and thus allow it to penetrate newmarkets,� says Russ Daniels, chief technology o�cer for HP’s cloudservices strategy.
Hop, skip and jumpThe mobile phone has already enabled developing countries to skip �xedline networks. Cloud computing could prove to bea similar �leapfrog� technology because itdispenses with the need to build a cumbersome IT infrastructure. �Software developers from a developing country can buildjust as great an application on our platformas somebody who lives in Palo Alto,� saysMr Benio� of Salesforce.
Indeed, countries such as India will certainly take a big chunk of the global marketfor cloud services. Zoho, a popular suite ofwebbased applications, is operated by anIndian company, AdventNet. Indian hospitals are already o�ering specialised healthcare services this way. The insurance armof ICICI, an Indian bank, has used thetechnology to come up with innovativeservices such as a personalised insurancefor diabetes. Premiums are adjusted depending on how well policyholders stickto a �tness plan. All this suggests that theeconomic impact of the cloud may be feltnot only in the IT industry itself but in other sectors too. The internet disrupted themusic business; Google disrupted the media; cloudbased companies could becomedisrupters in other ine�cient industries.Buzzwords such as �Health Care 2.0�,�Banking 2.0� or even �Education 2.0�could soon acquire real meaning.
But given that the cloud is global by nature, how might it be regulated? The conclusion of this special report will o�ersome answers. 7
8Dangers within
Source: Forrester
Research
How IT organisations feel about risk ofemployee-driven, unsanctioned use of Web 2.0*June 2008, %
*Survey of 242 US IT decision-makers
at companies with 500+ employees
0 10 20 30 40 50
Very concerned
Somewhat concerned
Somewhat unconcerned
Completely unconcerned
14 A special report on corporate IT The Economist October 25th 2008
2
Previous special reports and a list offorthcoming ones can be found online
www.economist.com/specialreports
Future special reportsCountries and regionsSpain November 8thRussia November 29thIndia December 13th
Business, �nance, economics and ideasCars in the developing world November 15th The sea January 3rd 2009The future of �nance Jan 31st 2009
www.economist.com/rights
O�er to readersReprints of this special report are available at aprice of £3.50 plus postage and packing. A minimum order of �ve copies is required.
Corporate o�erCustomisation options on corporate orders of 100or more are available. Please contact us to discussyour requirements.
Send all orders to:
The Rights and Syndication Department26 Red Lion SquareLondon WC1R 4HQ
Tel +44 (0)20 7576 8148Fax +44 (0)20 7576 8492email: [email protected]
For more information and to order special reportsand reprints online, please visit our website
that will arise with the cloud. In one way itis the ultimate form of globalisation: vastvirtualised computer systems and electronic services know no borders. Yet governments are likely to go to great lengths toavoid losing even more control.
When the internet went mainstream inthe late 1990s, libertarian thinkers arguedthat cyberspace was a distinct place callingfor laws and legal institutions of its own.After all, they said, it was built in such away �that it interprets censorship as damage and routes around it�. But many governments quickly found ways to blockcontent they deemed o�ensive. Just look atChina and its �great �rewall�.
Controlling where data are stored andhow they are treated is harder, though, because information can �oat freely in thecloud. And it is not just undemocratic governments that want to control their citizens’ and companies’ data: indeed thereare nearly as many sets of data regulationas there are countries. �If we wanted to beon the safe side in terms of regulation, weprobably would need 95 individual datacentres,� says Chuck Hollis, a technologistat EMC, the leading maker of storage gear,which owns Mozy, a cloud service that allows users to back up their data.
There are technological �xes to this problem, too. Customers of Amazon’s storageservice, for instance, can have their datakept either in an American or a Europeandata centre. In future, cloud providers willo�er many more options for where dataare kept and how they are protected. Itseems possible that data as well as content
will eventually travel with security, location and expiry policies attached.
It is when computers become virtualmachines that things get really tricky.These days IT systems are at the core ofmany companies�and just like data, thesesystems can now live in a variety of places.What happens if they start to migrate toanother country where power is cheaperor regulation laxer? Similarly, if servicesare a combination of elements provided indi�erent jurisdictions, who is liable ifsomething goes wrong?
The cloud’s potential political and social e�ect is only now entering the publicdebate. IT �rms are putting it on the agendaon both sides of the Atlantic. In midSeptember, for instance, Google organised adiscussion on the subject in Washington,DC. A few days later SAP presented a whitepaper in Brussels pointing out that policymakers are not aware of the dramatic economic impact of the �Future Internet�, asthe paper calls the cloud.
IT industry leaders note that o�cialsfrom many countries have begun to takean interest in the cloud. Some just wantdata centres to be built in their country tocreate jobs; others are concerned about issues of law enforcement and jurisdiction.The danger, they say, is that cloud providers might be obliged to build more datacentres than are needed and have to comply with many di�erent regulatory regimes. Some of them have been �oatingthe idea of �freetrade zones� for data centres where common rules would apply.
Yet such ideas appear at odds with another big question that could keep thecloud from growing: how to protect privacy. �Consumers expect their informa
tion will be treated the same way on thecloud as if it were stored at home on theirown computers,� says Ari Schwartz of theCenter for Democracy and Technology, anadvocacy group. Many of the devices thatfeed into it, such as sensors and cameras,will be intrusive. For example, Google’svans go round taking streetbystreet pictures of cities for the company’s onlinemap service called �Streetview�. The pictures are meant to help people �nd theirway around, but also often show passersby in embarrassing situations. To protect their identity, Google now blurs theirfaces and licence plates.
Hands o� my dataBut it is not only personal information thatcould get out into the open. Privacy is aworry for companies too�and not just because criminals or spies might intercepttheir data. Once they are in the cloud, governments can also get their hands on themmore easily. SWIFT, the organisation thatmanages international bank transfers, isplanning to build a data centre in neutralSwitzerland. That will allow it to keep dataabout European transfers on the old continent, where it cannot be subpoenaed bythe American government. Webbased email is not safe either. Thanks to the StoredCommunications Act, American law enforcers can read people’s messages�anddo not even have to tell the recipient.
Just as too much regulation may keepthe cloud from rising high, so could lack ofprivacy. If consumers and companies cannot be sure that their information is safe,they will err on the side of caution. But despite all the caveats, the precipitation fromthe cloud will be huge. 7