As an industry, we've been looking at cloud computing benefits in too shallow of a way. If the cloud is truly revolutionary, then it has to be able to support models of productivity and user experience that aren't available to customers today. That means the real cloud of the future must be what we can call a "super cloud," which should be a platform that makes computing and networking better -- not just cheaper. The benefits of that kind of cloud could be twofold for service providers, increasing IT spending and the value to IT shops.
But here's the big question: Can cloud providers build it?
Basic cloud computing is a game of balancing costs and economies of scale. Enterprises run applications today in their own data centers, and the cloud's proposition is that these apps can be run for less money in the cloud while creating profit for the cloud provider. You don't have to be a financial wizard to realize this kind of cost-based goal is going to be hard to achieve. The return on investment (ROI) for a cloud project has to come out of pure cost savings, and any perceived risks of cloud migration force buyers to demand higher ROIs -- all of which makes the justification for cloud services more difficult. No wonder we have less than 2% penetration of cloud computing, as measured by its share of total IT spending.
But instead of shaping their message entirely around cost reduction, providers should focus on expanding cloud computing benefits for enterprises. This is the essence of the "super cloud" concept. For a cloud provider, it means offering users better personal experiences or higher worker productivity than they could obtain through traditional IT models. Since ROI is the ratio of benefits to costs or risks, raising the overall amount of benefits raises ROI and increases the number of cloud projects or services that customers can justify.
For a super cloud to offer better experiences and services than what traditional IT models provide, it has to support applications and features in a way traditional IT can't. Cloud providers are already expanding their services from very basic virtual machine hosting, known as Infrastructure as a Service (IaaS), to higher level Platform as a Service (PaaS) and other cloud platform offerings. Amazon Web Services offers a cloud-based content delivery network service, for example, and Alcatel-Lucent's CloudBand product defines "carrier PaaS" application programming interfaces (APIs) to help cloud operators build special application components and features that not only help control costs, but also expand the cloud's utility.
NFV as a path to the 'super cloud'
In the super cloud, basic cloud services -- IaaS, for example -- are augmented by feature components that could represent application components, hosted network service features, or even network features exposed via management APIs. Developers could use this combination of platform tools to create applications and cloud services, just as they build applications today using tools exposed by traditional operating systems and middleware. In that sense, the super cloud can be thought of as a distributed, "super" operating system -- something that would attract its own development community.
Network functions virtualization (NFV) may be the way to create and deliver more PaaS and super cloud features. While NFV doesn't mandate cloud hosting of virtual functions, it certainly supports it, and most NFV implementations announced so far combine a mix of cloud application components and network functions into a single service. That creates what could be the starting point for the super cloud because it could blend location, mobility and social networking information with applications, network delivery and network connectivity. If cloud providers could tap such a unified ecosystem to quickly adapt services to market needs, they would raise profits quickly and drive much faster cloud adoption.
But this raises a question about how virtual functions and software components hosted in the super cloud would connect to cooperative features or services. Every software component has at least one interface, and to create cooperative systems of components, you have to link these interfaces into a pattern of workflow -- a process sometimes called orchestration. NFV implementations may include orchestration capabilities, and there are various software orchestration tools that could work with them. For example, service-oriented architecture (SOA) defines workflow orchestration. Meanwhile, standards groups such as the Object Management Group and TM Forum also have assets that cloud providers could harness to define super cloud PaaS orchestration. The challenge is less about determining whether there's a standard to adopt this model and more about deciding which standard would be best.
Building the super cloud: Challenges remain
This does lead to a potential tradeoff, however, which the super cloud must address. The flexibility and agility needed to address market opportunities may conflict with the stability and standardization that's required to ensure low operating costs and effective customer support. A single set of super cloud platform APIs would be attractive to developers because it would allow them to create applications for any super cloud ecosystem, but the very concept of the super cloud is only now emerging and so it might be difficult to pick the right tools at this early point.
Operating systems and middleware present developers with a consistent toolkit for building applications. The super cloud may have to answer the question of whether this stable and static mechanism is appropriate for building cloud-hosted experiences in the future. There are software design techniques -- SOA being the most visible -- that could let developers browse inventories of service elements and select those that fit for integration into a service workflow. None have been tested at the scale of a cloud provider's production environment, and none offer the agility that the super cloud demands. Will new models emerge? That's a key question.
Management is likely the final question that the super cloud concept will raise. Virtualization is already stressing current management models by eliminating the simple notion that resources are "real" and can be classified according to the type of device. How are virtual resources classified, and how do dynamically composed experiences look to a management system? The industry has been nibbling at the edges of this problem for half a decade -- since virtualization deployments began -- and still have not fully resolved it even for the limited data center model. The super cloud will have to be "super" everywhere, including management, or it will be unable to support its dual missions of satisfying user needs and delivering carrier profits.
About the author:
Tom Nolle is president of CIMI Corp., a strategic consulting firm specializing in telecom and data communications since 1982. He is also chief architect of the CloudNFV project, a TM Forum-based implementation of the European Telecommunications Standards Institute's model for network functions virtualization (NFV).
What cloud architecture would support the next-generation cloud?
Renewed focus on cloud security in 2014
Cloud market forecast in 2014