As the latest release of the open-source cloud infrastructure debuted, controversy swirled anew. Will OpenStack kill third-party PaaSes or vice versa?
OpenStack turns 3 years old — to mixed reviews; Dell deal in deep doo-doo; and more cloud computing news from around the interwebs.
In cloud computing, “the stack” typically refers to a simplified three-tier view that divides the cloud’s components into infrastructure, platform or software/applications. While conventional wisdom suggests that vendors constantly try to move up this stack, from the lower margins of infrastructure toward the higher margins and customer loyalty of services/applications/software, companies such as Amazon, Salesforce and VMware are increasingly trying to dominate several points at once.
Amazon Web Services began at the bottom of the cloud stack by offering basic storage and basic virtual machines, with price a key part of the proposition. Third parties like RightScale emerged, layering value and functionality on top of Amazon’s bare bones, and as Amazon added features and began to move up the stack, dependents such as RightScale were forced to diversify in order to remain relevant.
Toward the top of the stack, Salesforce began by offering a hosted CRM application. The company has subsequently diversified down the stack, releasing and continuing to enhance the Force.com platform upon which a growing range of third-party developers build and sell their own applications. In parallel, the Salesforce CRM application has continued to gain functionality, often adding features that challenge vendors selling products on top of Force.com.
Similarly to Amazon, VMware started with infrastructure, developing a virtualization solution for desktops and servers. The company has diversified greatly since then, with this week’s addition of Horizon App Manager leading cloud computing analyst Ben Kepes to remark that “[VMware does] the entire stack now.”
These companies, among others, are diversifying their products for a wide variety of reasons. The Leading Edge Forum’s Simon Wardley describes the approach adopted by both Salesforce and more recently VMware as a classic example of a “tower and moat” defense, in which the “principle here is to defend a revenue stream (the tower) by creating a moat devoid of differential value with high barriers to entry around it.” In other words, Salesforce prevents the competition from developing viable challengers to its revenue-generating CRM products by filling the space around them with open-source or low-revenue products. There is no easy way for potential competitors to get close, especially when those competitors are companies like Oracle and Microsoft, which are more accustomed to the very different revenue models associated with on-premise installed software.
Amazon, Salesforce and VMware also work hard to build a network of complementary products and solutions in which the whole is far more compelling than each of its parts. Amazon, for example, began by selling access to commodity storage with S3. This area of the market is one in which anyone can compete, as one source of online storage is typically much the same as any other. Amazon keeps its storage offer compelling by means of reasonably low pricing and by surrounding it with a range of other web services such as EC2. Individually, none are really much better than those offered by Amazon’s competitors. But together they offer a proposition that is hard to beat without replicating everything Amazon already offers and in the process incurring huge costs.
As the companies that underpin the cloud continue to grow, the range of products they offer will continue to expand. Some of those products provide new revenue-creating opportunities, while others are simply there to squeeze the competition. However, as the ecosystem (value-adding partners of Amazon such as RightScale, revenue-generating customers of Force.com, etc.) itself grows in scope and value, a long-visible tension will become increasingly serious for all concerned. Every feature added to Amazon Web Services as it competes with Rackspace and other cloud infrastructure companies is potentially a feature that encroaches on the service offered by partners like RightScale. Can Amazon grow its own capabilities at the same time as it nurtures an ecosystem of partners? Can any of these companies pursue their original product strategies while still remaning platforms upon which others can build and rely? Whether they must become one or the other remains to be seen.
Question of the week
A battle rages for the hearts, minds and wallets of cloud computing’s users. Public, hybrid and private clouds each deliver clear value for today’s enterprise customers, but it can’t be many more years before the public cloud’s ubiquity and economies of scale win over alternative models. By now, it’s a question of when — not if — the vast majority of computing will be performed on infrastructure that’s owned and operated by a range of public cloud providers.
The Leading Edge Forum‘s Simon Wardley published a long blog post this week that built on his earlier work about the rise of utility computing. In this latest post, he unpicks the various ways that cloud technologies are used inside the enterprise firewall. These technologies are typically referred to as private clouds, but Wardley argues that private cloud actually addresses only one of two very different use cases, and that two separate labels are therefore required in order to describe what is going on. He calls one use case “private cloud” and the other “enterprise cloud,” and goes on to explore each in more detail. According to Wardley, a private cloud is essentially the same as a public cloud like Amazon’s, but behind the firewall and dedicated to a single customer. It is created in response to the same drivers as a public cloud, seeks similar economies of scale and is essentially based on the requirement to be “good enough;” it is simply a commodity.
Wardley’s new concept of the enterprise cloud, on the other hand, is seen by him as a logical evolution of the philosophy that traditionally drove enterprise IT deployments. Altering this underlying belief, and re-architecting legacy applications in order to fit a more traditional cloud model, Wardley argues, would prove prohibitively expensive, preventing more significant adoption of public cloud techniques in addressing these use cases.
While I agree broadly with Wardley’s arguments, I would question whether they lead to his conclusion of two wholly different visions for cloud inside the enterprise. They appear to be two extreme ends to a spectrum of possible use cases, and shouldn’t prevent the application of mainstream cloud methodologies inside the enterprise. In both public and non-public clouds, there are valid business reasons to deploy infrastructure that is either cheap or robust, more fault tolerance or less, faster or slower, or whatever else may be required. Increasingly, we’ll see cloud providers explicitly differentiate products using a number of criteria. Alongside the bargain basement utility cloud, you’ll be able to pay a premium to specify the country in which your data will reside, the degree of fault tolerance you are prepared to accept in equipment, the bandwidth into (or out of) the data center, etc. The basic cloud is getting cheaper, but that doesn’t remove the opportunity to differentiate and charge for premium capabilities.
Wardley ends by suggesting that his enterprise cloud will become increasingly targeted toward a diminishing number of niche requirements in markets such as Defense, while public clouds become “the dominant form.”
Elsewhere this week, RedMonk‘s James Governor wrote about the shifting balance between the public and private cloud. He cited figures from an EMC presentation, which claimed that both public and private clouds will grow significantly over the next decade at the expense of legacy enterprise IT. The diagram Governor used suggests 70 percent legacy IT in 2010, falling to just over 30 percentby 2020. Over the same period, public cloud grows from about 20 percent to about 35 percent, and private cloud grows from around 10 percent to around 30 percent.
Governor expressed some surprise at the figures, and I agree. EMC’s premise that legacy IT will not disappear overnight is clearly correct, as is a prediction that private cloud will grow in the short term. However, I fully expect private cloud to be in terminal decline by 2020, as today’s arguments in its favor are overturned. Many of the barriers (such as security) to public cloud adoption are perceived rather than real, created by genuine misunderstanding, poor explanation or deliberate confusion sown by incumbent vendors. Others, including the need to maximize existing investment in recently constructed enterprise data centers, are of only short term relevance.
By 2020, individuals, SMEs and global corporations will all be placing the majority of their data in the cloud — the public cloud.
Related Research: Defining Internal Clouds: From Appistry to VMware