Although cloud computing has gained an enormous following, it still garners only a small fraction of IT spending....
To increase cloud usage, cloud computing has to drive changes in how companies build and use applications. Those changes will focus on the intersection between cloud components and legacy applications -- the hybrid cloud. The most important trends in hybrid clouds are the evolution from compartments to components, the emergence of "migratory processing" and the movement toward event-based applications.
Nearly all the early cloud applications that enterprises adopted were forms of hosted server consolidation. These applications were logical starting points because they used hardware and support resources very inefficiently, and they were largely independent within the businesses' overall IT processes. A departmental server running in the cloud serves the department, as before, so the impact of integration is minimal.
The whole notion of a hybrid cloud emerged because it was clear that further expansion of cloud computing would involve applications that were more connected with mainstream IT, so information flows across the public-cloud-to-data-center boundary had to be supported. This process begins with basic integration of workflows, but businesses are now realizing that if work can be exchanged across the public cloud boundary, then cloud resources can back up traditional data center assets.
Envisioning components rather than compartments
In this new hybrid cloud model, applications are viewed as components rather than as fixed silos of software and hardware. As a result, integrating the cloud with the data center has become the process of creating inter-component workflows that don't care where the components reside. Cloud-specific integration and even cloud bursting are going to be fully integrated with workflow management and APIs, and they will be validated using traditional application lifecycle management (ALM) processes.
Incorporating integration and connection of cloud components into standard workflows and ALM leads to another trend, the emergence of so-called migratory processing. Hybrid cloud evolution is inhibited if special practices are needed to use cloud resources, so the goal of workflow and API design is morphing from accommodating the cloud to making the cloud transparent.
Most users will get enough server efficiency from virtualization that they won't need private clouds. If that were the only variable, private cloud usage would remain low and most hybrid clouds would be integrating public cloud services with non-cloud data centers. Migratory processing demands that application components be transportable to any resource, which means an architecture that makes all resources look the same is critical. This can only be done by building internal IT on cloud-management tools.
Resource independence at the component level has a significant effect on application performance management and monitoring. Historically, users have managed server pools as "liquid capacity" rather than looking at the specific component-to-resource relationships. That's harder to do when some resources are in the public cloud, and particularly difficult if one goal of moving components is to meet specific quality of experience goals set to ensure worker productivity. Some believe that cloud orchestration tools and DevOps will take up a greater level of management integration, but it's likely that application performance management will be more complicated for some time because even basic practices and policies are in their formative state. Only greater experience in the hybrid cloud will create an evolutionary path that enterprises will trust and adopt.
Moving toward event-driven applications
The management integration issues for the hybrid cloud are mirrors of a larger trend, which is toward event-driven applications rather than traditional workflows. One benefit of the hybrid cloud is greater responsiveness to demands to improve worker productivity, particularly through the support of mobile workers and point-of-activity empowerment. Experience with mobility even in the consumer space has demonstrated that mobile workers are likely to be more reactive than proactive -- they need answers to questions and problems presented to them by the context of their jobs. That means software has to respond to worker events, not drive worker behavior.
Event-based applications could be the largest change created by the hybrid cloud. While event-driven programming is common in communications (such as protocol handlers) and manufacturing and process control, it's rarely part of general business computing. The challenge with event-based applications, wherever they run, is that it's necessary to maintain context almost on a per-worker basis. Properly done, this can enhance security, performance and productivity, but it will almost certainly have to evolve into existing application architectures from the Web or from the front end inward, or the changes demanded of planners, architects and CIOs will be difficult to absorb.
Hybrid cloud trends, and all other cloud computing trends, are driven by the parallel evolution of productivity and empowerment mechanisms and the demand for high-level resource efficiency and performance. The former trends move IT toward componentization, per-worker customization and worker event-driven information movement. The latter trends move us toward virtualization, cloud computing, hybrid resource pools and resource independence. Neither of these trends can be realized without progress being made on the other. The hybrid cloud is important, not because it's the ultimate goal but because it's a step that brings the two trends close enough that planners can grasp a little of each and exploit them to advance the evolution of IT within their organizations.