社長ブログ：クラウドビジネスに注目！

More than a year ago in the US, the financial crisis, along with a sudden spike in gas prices the summer before, led  many consumers to stop buying cars. The recession that started then, in addition to increasing public awareness that excessive fossil fuel consumption is leading to global warming, has led many drivers to stop driving and to seek alternative transportation, such as car pools.  Gas guzzling SUVs are being replaced, or simply not driven any more.  Public awareness on climate change is leading to major changes in the business environment.  In this blog, I will argue that in the not so distant future, these changes will also have a major impact on the IT industry, in particular on server computer and server application vendors.

In the US,  GM and Chrysler, have in the past fifteen months gone through a period of drastic reductions in sales and enormous losses.  Earlier this year these two companies filed for bankruptcy, and now they are using government funds to reinvent themselves.  This year these  makers have started sales of hybrid cars and are planning to launch electric cars in the near future.  Some consumers are responding well to the hybrid offerings but many are waiting for the electric cars.  Hybrid vehicles may turn out to be a transitional solution on the way to electric vehicles which are much more efficient with a much smaller carbon footprint.  As the car industry remakes itself to produce electric cars, it is unclear whether the present players will make the transition.  Some will succeed, others will not.

In many cities in the US private cars are the only way to commute to work.  By comparison, transit systems such as the JR train system in Japan are utility systems shared by tens of millions of users each day in all major cities. Each user only pays a fee for the distance traveled. When comparing the amount of CO2 generated per passenger when commuting to work, the train achieves much higher efficiency than the private car, resulting in a much smaller carbon footprint.  However, when private cars are used for the “The Last Mile” to access the train system or in the country side where the train system may be unavailable, they become part of  the transit system  and the overall carbon foot print is small by comparison.

Worldwide services like Google,  SalesForce.com and Amazon.com are also systems shared by hundreds of million of users non-stop each day.  They  are  based on massive infrastructures with fees based on actual usage.  They are the equivalent of the JR Transit System.  Because of their massive scale they can be called IT utility clouds.  The IT industry is also about to be affected in a major way by the new business environment that is being brought about by climate change.  In the comparison below I will argue that the rising cost of electricity,  as well as public awareness of the environment, are going to result in a major switch by corporations to the utility clouds.

Public awareness of  the risks of  climate change is increasing worldwide, and  this month’s United Nations Copenhagen conference (COP15), or perhaps later agreements, will most likely create a cap on the amount of CO2 emitted per country, per industry and eventually per corporation.  Those companies exceeding their quota will have to pay a tax or buy credits from companies that haven’t reached their emissions cap.  Electric utilities create electricity by burning coal and are major CO2 polluters, thus they will face steeply increasing costs as cap and trade schemes kick in.  Because of COP15 or later measures, the price of electricity is bound to rise steeply.  In the US, this is expected to bring a price increase of up to 30% in the not so distant future.  Higher electricity prices will drive corporations to reduce electricity consumption by eliminating unnecessary use.

High on the list of electricity guzzlers are corporate IT servers which  must run non-stop 24 hours a day,  even though they are in use only a fraction of the time.   Corporate IT servers are obvious targets for replacement.  A fundamental problem with corporate IT servers is that they run only one application in single tenant mode and thus IT corporate servers leave a large carbon footprint when compared with cloud utilities.  However,  as with  private cars,  when corporate IT servers are used to access utility clouds or to run applications that are not available yet on the cloud  they become part of the cloud and the overall carbon footprint is small by comparison.

As companies stop buying or using IT servers, the makers of these servers are going through the same wrenching process that car makers have been experiencing since last year. With looming high electricity prices on the horizon, some of these makers have been trying to reinvent themselves as virtualization service providers. Virtualization software enables server hardware to be operated in multi-tenant mode, vastly increasing efficiency.  Still, these so called private clouds continue to operate legacy software are used only a fraction of the time.  In public clouds, the sharing of the applications increase as multiple corporations share the server.  However,  in either case, virtualization clouds are much less efficient than utility clouds because they use legacy software.  Virtualization clouds are in a sense transitional solutions as enterprise applications are moved or remade into utility clouds.  As with corporate IT servers, when virtualization clouds are used to access utility clouds they become part of the cloud utility  and the efficiency of virtualization clouds greatly increases by comparison.

The new business environment will have also a major impact on the makers of corporate IT server software.  Microsoft and other legacy software makers are about to go through same changes as users stop buying and deploying the IT server software that brings in annual revenues in excess of tens of billion dollars.  Hardware makers are trying to transform themselves into virtualization cloud providers, so what about legacy software makers?  Microsoft is trying to become a cloud utility provider.  Next year,  Microsoft will start its new BPOS cloud service.   Can a company that has for fifteen years developed and licensed package software remake itself as a cloud provider?  Can a car maker remake itself as a transit utility?

After  fifteen  years  of client-server computing and the proliferation of hundreds of millions of energy inefficient IT servers, the advent of Cloud computing is a welcome development for our planet.  The companies that made enormous profits from client-server computing must reinvent themselves.   Some of the major players will survive, others will not.

Nicholas Carr’s book, “The Big Switch”, compares the current struggle between Client-Server and Cloud Computing to a similar struggle that took place more than one hundred years ago within the electric power industry. Thomas Edison, the great inventor, believed that electricity must be generated locally by every company or factory using generator equipment made by his company, General Electric, while his adversary and former colleague, Samuel Insull, believed that electricity generation was the job of a central utility, run by his company, Chicago Commonwealth. After a difficult and long fifteen year war between them, Insull won and Edison abandoned the generator manufacturing business. In today’s blog, I will discuss Carr’s analogy and explain why Cloud utility companies like Google, Amazon and Sales Force will win over Client-Server vendors like MS, IBM, etc.

When Insull took over Chicago Commonwealth in 1893, his strategy was to become the lowest cost producer of electricity and to lower prices to levels that competitors could not match. He was able to achieve this goal by bringing economies of scale to his production of electricity. Companies and factories that bought and operated Edison’s generator’s to provide electricity to its users had to operate a 24 hour service despite of the fact that electricity was consumed only a few hours a day. In modern terms, this is much like companies hosting their own servers to provide common IT services.

Insull recognized that the load factor in companies operating their own generators was too low and therefore set out to achieve higher load factors in his own operation. To do so, Insull lowered prices to attract companies from as many industries, in as many sizes, across as many areas as possible within the City of Chicago. At the same time, in order to achieve the lowest cost per kilowatt-hour generated, Insull strove to acquire and deploy the largest generators in the world, analogous to Cloud Data Centers in today’s struggle. Since Insull was able to achieve the highest load factor while operating the biggest generators, he was able to generate electricity at the lowest cost and was able to price his service much lower than the cost to companies and factories operating their own local generators.

Insull’s lower prices attracted increasing number of users, allowing him to lower prices even further. The resulting virtuous circle, from increasing users to decreasing prices, made it only a question of time as to when Insull’s utility would succeed. By 1918, more than fifty percent of the power consumed by factories and corporations in the City of Chicago was provided by utilities.

In my previous blog I described how Google, like Insull, is operating the biggest data center complexes in the world and Google’s share of the US and Europe Internet search market exceeds 67%. Utilization rates are not available for the Google Apps service, but since they are using the same servers for internet searches, we can safely assume that their load factor is pretty high, higher in any case than those for companies running their owns servers. Google is therefore able to price the Google Apps service much lower than what it costs corporations to operate their own servers.

It might seem unfair to compare the Google service against those of companies operating their own servers. How about comparing Google Apps and Microsoft Business Productivity Online Suite (BPOS) services? Due to their late start, the scale of Microsoft data centers is much smaller than that of Google’s planet wide computer (October Blog), but even if their scale were the same, their load factors would be significantly different. This is because the market share of Microsoft’s Internet search services is less than one tenth of Google’s share. Additionally, BPOS has probably a much smaller fraction of Google Apps’ share in the market for software as as service. This means that every day, Google serves ten times more queries than MS and therefore has a higher utilization rate for its data centers. Higher efficiencies lead to lower prices which lead to increasing numbers of users. It will only be a matter of time before the struggle is settled.

Will it take fifteen years, as in the struggle between Insull and Edison? A critical factor in the operation of a data center is efficiency in operation, in the simplest terms, this means turning out the greatest output for the lowest cost. Electricity cost accounts for most of the variable cost of operating a data center, and while the price of electricity for corporations has not changed much in the last 100 years, the explosive increase in the need for electricity in the last decade and the resulting uncontrolled burning of coal has created a planet wide global warming crisis. Since the price of electricity is bound to dramatically increase to reflect its impact on climate change, it is likely that the above struggle will be settled in much shorter time.

Recently a number of companies have announced cloud services to compete with Google Apps. Microsoft and IBM are among those that have recently announced their own cloud services. In the confusion resulting from these announcements, the end user must know that not all cloud services are created equal.

In today’s blog I will try to clarify why.  In my next blog I will discuss Cloud Security.  Firstly, I will emphasize that Cloud Services, Google Apps, Amazon EC2, SalesForce (in this blog for simplicity I will concentrate on Google and Google Apps) are based on a planet-wide, distributed, massively parallel computers that appear to the end user no different from the corporate server running, say an e-mail application like MS Exchange.   Secondly, I will emphasize that cloud services are a disruptive innovation that will be very difficult for existing client-server computing vendors, like Microsoft and IBM, to successfully introduce cloud services as a business, even if they are able to create a planet-wide platform to rival Google’s.

The Cloud is a planet-wide, massively parallel computer

Google’s planet-wide computer has been evolving for the past 7-8 years as the company’s share of Internet searches has risen to over 60%. Throughout this period, Google has gradually built a massively parallel computer to process within a second, each of the hundreds of millions of search queries that it must answer each day.  The architecture of the massively parallel worldwide computer is comprised of both software and hardware.

State of the art software (Big Table, GFS, Map reduce) distributes and processes these queries in parallel as a single non-stop, always available computer. This software is based on decades of computer science research on massively parallel processing as well as fault tolerant experience acquired by servicing million of user queries daily.

The hardware platform is built on hundreds of thousands of custom-built PC servers distributed across dozens of data centers.  Each data center is estimated to require around $500 million to build.  These data centers have been built up gradually over Google’s lifetime.  During this period, Google has developed proprietary technology to package, house and cool hundreds of thousands of processors in one data center.  Google’s data centers are said to be among the most eco-friendly, energy efficient data centers in the world.

Disruptive innovation

High Technology products frequently emerge to satisfy market needs, displacing old inefficient products.  Innovation drives the evolution of technology to satisfy market needs.  Most frequently, innovation occurs incrementally and progresses as technology and user needs evolve.   This type of innovation is called sustaining innovation.  A new version release of the MS Office or MS Exchange is typical examples of step-ups in the sustaining innovation curve.

However, over the years, products driven by sustaining innovation climb up a curve that increasingly diverges from that of the user needs curve, which typically rises at a gentler pace. Subsequent releases of the MS Exchange product have cumulatively loaded the products with a myriad of features that the typical user has increasingly had little use for.  This MS Exchange feature overshoot exemplifies the divergence driven by sustaining innovation between the evolution of a product and the needs of the user.  While the dominant vendor’s product curve moves away from the user needs curve, a different product (e.g. Google Apps) creeps up on a trajectory below that of the user needs curve but at some point, crosses it, emerging to displace the established product.

Web based ASP services for the enterprises, available almost from start of the Web, have been unable to meet user needs until the last couple of years.  However, the spread of broadband and the evolution of Ajax and the Web 2.0 platform have, since roughly last year, suddenly allowed Web ASP services to match the needs of many corporate users.  Google Apps is able to offer the core features provided by the MS Exchange applications and at a much lower cost. Google Apps is now poised to replace MS Exchange for enterprise users.

Given the threat that cloud services represent to established vendors such as Microsoft and IBM is it possible for them to succeed in creating their own cloud services to compete?  In ‘The Innovator’s Dilemma’, Christensen argues that it is difficult for companies that have been advancing their products on a sustaining innovation trajectory to move to a competing disruptive innovation curve.  The main reason being that making the shift means adopting values and processes that are not compatible with the existing culture within that corporation. Microsoft derives very high profits from MS Exchange and high profit deals are a core value within the corporate sales culture.  Thus, it will be difficult for the high gross profit Microsoft package sales culture to adapt to the new low profit cloud services culture.  Christensen mentions that existing companies were only successful when they were able to spin-off the disruptive technology products to a newly created and completely independent organization within their structure.

Typically, established vendors will try to retain the high revenue/high profit of their established products by offering hybrid solutions, in this case a client server/cloud solution, but it will be difficult for the hybrid solutions to compete in cost performance against the purely disruptive innovation such as Google’s Cloud.