Commerce, it's said, is a fair-value exchange between buyers and sellers. It's easy to forget, when we look at something as exciting and revolutionary as the cloud, that it's still subject to the rules of business logic. Infrastructure as a Service
Lately, the answer seemed to be Platform as a Service (PaaS). PaaS pulls more software elements into the cloud by adding an operating system and middleware, thus displacing more software and support costs and adding potential value. The problem is that Microsoft, with Windows Azure, had the only viable PaaS offering -- but Azure was never positioned or developed as it needed to be to establish PaaS. And what Amazon is now creating is leaving Microsoft's PaaS at the station. It's not IaaS, but it's something that goes beyond IaaS: platform services.
Platform services pull the high-value application elements out of traditional OS and middleware and make them into Web services that can augment any application running on IaaS. For the cloud provider, this means a new source of revenue. For application developers, it means building additional value for their applications without requiring custom coding for the cloud. For users, it means increasing the flexibility and utility of the applications they run.
Platform services often duplicate commercial and open source software, but they're offered in "as a service" form and with support, ideal for users with a real commitment to the cloud. Database services are an example of platform services, as well as tools for deploying and managing cloud applications.
AWS takes platform services to the next level
In the wake of Microsoft's failures with PaaS, however, reigning cloud leader Amazon Web Services (AWS) has taken platform services in a new and exciting direction -- one that promises true cloud-specific applications. This would represent a true shift in cloud computing strategy, and could kill PaaS as we know it.
From the beginning of the cloud market, cloud providers have offered some form of database services as a platform service, because traditional database middleware typically requires data be stored locally with the application. Even private cloud stacks, such as OpenStack, offer a range of database services -- for OpenStack, ranging from simple block storage through the new Trove Database as a Service.
Amazon's take on platform services represents a profound shift in cloud computing, if a largely unrecognized one.
Amazon has offered these services, too, but it has quietly expanded its range of platform services, focusing beyond accommodating the cloud in applications to actually supporting apps specifically for the cloud that wouldn't be practical elsewhere.
What's different about Amazon's approach is it doesn't tie the user to a specific operating system/middleware combination as PaaS would, so it's possible to develop AWS-based applications for any platform, and in most cases, add cloud features even to existing applications with only modest efforts. Amazon's platform services include tools to facilitate deployment -- CloudFormation, Elastic Beanstalk, CloudWatch, OpsWorks -- but also cloud-specific application services, such as the recent AppStream and Kinesis, as well as virtual desktop, caching and even Web optimization services.
Amazon's take on platform services represents a profound shift in cloud computing, a largely unrecognized one. There are two immediate impacts of the movement toward platform services by market leader Amazon: cloud service momentum shifts decisively to IaaS and away from PaaS, and IaaS providers are pressured to create their own platform services that match and even better those of AWS.
The shift in cloud emphasis and opportunity from PaaS to platform services undermines the ability of any operating system or middleware vendor to leverage their current data center incumbency by providing a cloud version of their current platform. To stay even with the market, such a vendor -- including and especially Microsoft -- may have to present their current middleware as a series of discrete Web services rather than as a complete platform, making it easier to compete with.
PaaS, and many IaaS providers, could be killed by platform services because they pose a double risk. First, IaaS providers will be pressured to match platform services innovators like AWS, exposing them to higher costs. Second, Amazon's profits from platform services gives them more latitude to reduce the prices of IaaS, which would put competitors in a position of lowering revenues by matching the price reductions or losing market share based on price.
The risks of a cloud market shift to platform services
The shift to platform services is not without risks, though. There are no standard interfaces for platform services, and different IaaS providers may seek differentiation by using their own application programming interfaces (APIs), causing vendor lock-in. This means platform services users should be cautious about how they integrate the services into their applications -- particularly if they expect to use multiple IaaS providers or if their IaaS provider presents some risk of being acquired or going out of business.
Addressing the risk of an "API explosion" in platform services is critical as platform services become pervasive. The best strategy is to isolate application requests for platform services as much as possible from the rest of the logic so providers can be changed without a major impact. This can be facilitated by reviewing all of the implementations of a given platform service currently available, watching for new providers who offer it, and adapting applications gradually to conform to market conditions.
Amazon is far from done with platform services; they are likely only beginning their major push in this area. As they expand their offerings, they'll be demonstrating how to build cloud applications and not just move current applications to the cloud. That's a critical step in generating the most from a cloud commitment.
About the author:
Tom Nolle is president of CIMI Corp., a strategic consulting firm specializing in telecommunications and data communications since 1982.
This was first published in February 2014