Jevon's Paradox is the proposition that technological progress that increases the efficiency with which a resource is used tends to increase (rather than decrease) the rate of consumption of that resource. When that increase is not accompanied by any more than organic growth in resource management, it can result in a resource crunch. Simon Wardley of the CSC Leadership Forum has often pointed out that, whilst cheaper operations costs and reduced capital spending should signal a reduction in TCO, the truth is quite the opposite. This argument is rather interesting on its own, but I see the greater issue for Cloud to be resource starvation. There are several parallels which reflect that, despite observed reductions in TCO, the resource crunch has resulted in both increased costs and reduced service.
We have witnessed this effect in the mobile space, where mobile bandwidth has been consumed at an alarming rate with the emergence of data services. AT&T's iPhone service in the US the last few years has only highlighted this, with the network brought to its knees in Manhattan, San Francisco, Los Angeles, Chicago and other major metropolitan centres. Advances in the utilisation of mobile network resources saw their near exhaustion in less than a decade. This has been accelerated and exacerbated by the common practise of planning infrastructure after the fact. Other regions have fared better, such as South Korea, Japan, and Europe, yet at the price of resourcing and implementing broad infrastructure projects.
Will the Public Cloud be the next usage model that only increases resource consumption? The present rate of adoption, for example, by small businesses and social network games of Amazon Web Services, seems to suggest so. These are early adopters, yet the benefits of Cloud, together with these trends, are significant enough to strongly indicate only an increase in consumption.
Despite all the buzz to the contrary, compute and storage resources are finite, andnetwork resources continue to provide the principal bottleneck. Presently, dominant Public Cloud campaigns gloss over, or entirely ignore, this finite nature. The more compute, storage and network resources transform into utilities, the greater the exponentiation of consumption. Virtualization only accelerates the consumption rate, given the increase in server density. Without understanding this, increased Cloud adoption will be pursued with poor management of resources, and, resultingly, costs will increase to a point where they are uneconomical.
Resource management will grow in importance as inter-connectivity and cloud consumption increase. Although billing and metering have developed to a certain degree within Cloud service providers, this has not been approached in a manner that would truly manage the use of resources. Much has been developed to support this at the virtualization layer, yet not at the service layer. Resource management aims to address this gap, by enabling actions to be taken with workloads based upon service level agreements and policies, resource profiles and how they map to workload execution within resource pools and individual nodes.
This impacts both TCO and ROI. TCO can be maintained at a consistent level of efficacy through managing resource availability and constraints, to optimise the cost-vs-service ratio. ROI can be targeted and achieved through assurance of resource availability. Without resource management, costs can easily increase beyond efficacy, and pathfinding activities suffocated, as resources are heavily consumed and capacity exhausted.
Whilst a high-volume, low-margin strategy suits a number of business models, those looking to the Cloud to execute their mission-critical or highly-regulated processing need more than an optimistic RAS model to assure their operations. Tenant visibility into the Cloud (more on this in my next post) combined with fine-grained resource management will make this actionable.
