Right now, the three major public clouds (Amazon, Microsoft, and Google) seem all shiny and new, like many technologies seemed at some point in the past. Let’s see if we can learn from history and assess the risk of the public cloud’s becoming just another legacy platform.
Legacy Platforms of the Past
Our industry’s history has seen numerous computing platforms. To qualify as a “platform” for the sake of this analysis, the computing environment has to have been a major environment both for innovation and for deployment of a broad class of applications. Applying these criteria leads us to the following legacy platforms:
- The mainframe: The legacy platform that everyone loves to bash still executes a healthy percentage of the back-end business transactions in the world. In the heyday of the mainframe era, IBM actually had competition. Burroughs, Sperry Univac, NCR, Control Data, and Honeywell (collectively referred to as the “BUNCH”) all built and sold mainframe-class systems. But the mainframe became a legacy platform back in the 1980s, when minicomputers from vendors like DEC, Prime, Wang, Data General, and Datapoint offered customers a more attractive price point and, in some cases, better applications and a better user experience.
- Minicomputers: Minicomputers had a short reign in the sun. Once IBM introduced the PC in 1983 and other vendors introduced compatible PCs (remember Compaq?), minicomputers were eclipsed due to the PC’s ease of use and, more importantly, the innovation in software development that Microsoft made possible with the first standard cross-vendor operating system, DOS. If you do not remember DOS, open a Command Prompt in Windows and imagine that this is your entire operating system. Of course, if you use Linux, this will feel very familiar.
- DOS operating system: Not long after IBM and Microsoft agreed to standardize with DOS as the operating system, both companies began working to replace DOS with something new and better. Microsoft developed Windows, which is with us to this day. IBM brought us OS/2. By 1993, DOS was dead, replaced primarily by Microsoft Windows.
- IBM OS/2: For a brief period of time, IBM had a chance to replace DOS with what was then a truly superior operating system—the first desktop OS that offered memory protection and multi-threaded applications. But Microsoft had the attention of the application developers, and the wealth of applications available for Windows resulted in the less capable OS’s triumph. Microsoft nailed the coffin shut on OS/2 when it came up with 32-bit Windows and Windows NT (the precursor to today’s Windows) and refused to license the Windows 32-bit libraries to IBM, leaving OS/2 with no base of third-party applications. This taught everyone in the industry an important lesson: the best platform does not always win—the platform with the most third-party developer support often wins instead.
The Common Threads in “Legacy”
Several common threads can be seen in those platforms that had their day in the sun but have now been relegated to legacy status:
- A more cost-effective alternative emerged: Mainframes and minicomputers became legacy because they were not based on the Intel chip architecture. Intel’s high volume and low price point for CPUs drove all competing processor architectures out of the market.
- A new platform offered a richer set of applications: DOS and Windows both won over competing alternatives, not because they were technically superior, but because Microsoft put together a developer evangelism program that ensured its platforms had a more robust set of applications than the alternatives did.
- The new platform did a better job of catering to end users: Users tend to adopt the technology that meets their needs without concern for the considerations that drive technical people in the industry. Windows beat OS/2 and Unix simply because end users wanted the Windows applications. The users didn’t care that OS/2 and Unix were technically superior offerings.
- The locus of innovation from the ecosystem shifted to a new platform: The question of which platform to build something new on involves one of the most interesting and freest markets on the planet. Developers choose which platforms to support based on a variety of factors, including ease of development, ease of managing the platforms in production, cost, ubiquity, and most importantly, what new tools and innovations supporting vendors are delivering for the platforms.
Two of the early signs that a platform is becoming legacy can be seen when the third-party community ceases to innovate on that platform and the burden of keeping it up to date falls almost entirely on the platform’s vendor. The pace of innovation in our industry is so fast that no single vendor can keep up with it. Therefore, any platform or platform vendor is only as good as its ecosystem. It is due to this point that the modern versions of Windows are at risk of becoming legacy platforms, overtaken by Linux.
Will the Public Clouds Become Legacy Platforms?
Right now, for some situations and some workloads, public clouds have a tremendous cost advantage over alternative on-premises architectures. But as we explained in “Cloud Bambi Economics,” the public cloud may be neither the cheapest place to start your business nor the cheapest place to run your business. The public cloud is not the hands-down winner in terms of being the most cost-effective platform.
We all marvel at the rich set of services delivered by the public cloud vendors. It is certainly much faster to sign up for a set of services at Amazon than it is to evaluate ten or twenty different management products and deploy them in your own environment. But in providing this set of services, the public cloud vendors are setting themselves up as “one-stop shops.” On the face of it, this is extremely attractive. However, it puts the cloud vendors in the business of keeping their management stacks up to date with innovations in the middleware, development tool, and management software industries. In this sense, the public cloud vendors have decided to compete with the entire Sand Hill Road venture funding ecosystem. Look at what happened to IBM, BMC, HP, and CA when they promised customers that their frameworks could manage everything. The “big four” because “blind dinosaurs”; vendors of outdated products that could not be modernized to meet the new requirements of users.
Windows and Linux both became successful because they were embraced by ecosystems of developers who provided a continuous stream of innovations on those platforms. None of the cloud vendors have yet built such an ecosystem – and their current efforts to subsume innovations into their platforms may be their undoing.
What Will Replace the Public Clouds?
If the public clouds from Amazon, Microsoft, and Google are to become legacy offerings, then what is going to replace them? We cannot know at this time, but we do know that the answer will have to offer lower costs, better choice, better manageability, and a superior ability to deliver solutions with quality and agility. Business’s demand for functionality implemented in software is infinite. With infinite demand, we are very likely to see a continued supply of innovations—some set of which may well sink today’s public clouds. Ecosystems will be the key to whatever emerges, as ecosystems of cooperating vendors always win over monolithic platforms.
The shiny and new public clouds of today could easily become the legacy platforms of tomorrow. The history of one vendor “doing it all” is not pretty, and the public cloud vendors are dangerously wedded to this very path.