Cloud computing and virtualization promises to revolutionize the architectural principles of the data center. Shared resources enable efficiency, but ultimately the dynamism required to achieve such gains in efficiency will cause chaos in a variety of other functions throughout IT. The CIO is in for a rocky road unless a broader set of IT management vendors pave the way for a smooth ride.

The (In)accuracy of Forecasting in a Dynamic Environment

Organizations rely on the ability to forecast project costs and anticipated ROI in order to prioritize and set budgets for coming years. Many IT project management centers have turned in recent years to Portfolio Project Management (PPM), also known as IT Portfolio Management (ITPM), systems to assist in forecasting and prioritization, pulling data from asset management systems and project management systems to determine total costs of a project – from hardware through development, from deployment through retirement.

PPM, however, may not be capable of dealing with the dynamic, shared nature of resources in its forecasting scenarios. Nor is it necessarily ready to adjust associated costs based on CPU and memory requirements rather than dedicated hardware. In a shared environment the costs of hardware and other resources are also shared in what some may see as eerily similar to chargeback accounting methods used in the days of mainframes and LPARs. In order to provide the data necessary for a PPM to perform analysis and create possible scenarios on which managers and CIOs can set priorities and assign budgets the associated costs must be known or at least estimated.

The variability in a shared environment means that the capacity afforded by one server cannot necessarily be used to determine how much hardware is necessary to meet service-level agreements. The user capacity of an application on one server may not be the same on another, identical server because it is impossible to ascertain what resources will be available at any given moment in time. Bean_CounterVirtual management systems that allow the reservation of resources to be designated for specific applications’ use defeat the efficiency gains of virtualizing systems in a dynamic environment as there will invariably be resources sitting idle; wasted.

That begs a very important question: does your cost basis turn to bandwidth? To CPU cycles? To memory? To a combination thereof? To the number of virtual instances required to meet capacity? And when capacity is exceeded and new hardware acquired, to which project will the costs of acquisition be charged? Is that cost, too, shared across all projects and applications that might use the shared resources? Or will virtualization and shared resources become a neat IT accounting trick that allows project managers to assign base acquisition costs to a smaller project in order to keep the costs of larger projects down? 

There is no such thing as too much information

It will be nearly impossible at first to accurately forecast the total costs of an application across its lifecycle. Until the application has been deployed and executing in a virtualized environment it will be difficult to generate models based on historical consumption of resources and accurately tie real costs to projects. Too, the costs of one project in terms of shared resources necessary may not translate well to other, similar projects as their execution model and pattern of usage may differ in ways that lead to a different rate of consumption of resources.

It will be necessary to collect data from the entire infrastructure in order to more accurately manage your project portfolio and to provide the necessary historical data upon which future models are based. That data will come not only from the application and the OS, but from the virtualization management systems and infrastructure, as well. As even network and application delivery network infrastructure solutions evolve to supporting virtual resource provisioning that usage needs to be reincorporated back into the cost structures of individual applications and projects. The costs of deploying and maintaining an application in a virtual environment requires more supporting infrastructure across a broader set of IT functions and will therefore need to be incorporated back into project portfolios if intelligent decisions on projects are to be made moving forward.

Active Portfolio Management

PPM solutions will need to become more active, more integrated, and more aware of all the components necessary to deliver applications to users in a virtual environment. With the focus shifting from entire systems to components, e.g. CPU, memory, bandwidth, it becomes necessary for the processes and systems used to forecast costs and prioritize projects become capable of adjusting for such variables and incorporate them into its scenario generation and trending capabilities.

Vendors such as IBM and Compuware are well positioned to provide a more active project portfolio management solution as both have acquired PPM technologies and are also steeped in network and systems management implementations. The combination of PPM and NSM will be capable of collecting and correlating data collected from disparate pieces of the supporting infrastructure and provide the information necessary to more accurately model actual costs incurred by applications in such virtual environments.

CA and HP, too, have the means by which such data can be collected in near real-time, providing the means by which forecasting of costs can more accurately be conducted. The ability to forecast in a dynamic, continuous fashion will provide some of the data necessary to reduce the initial inaccuracies introduced by a shared resource infrastructure. But don’t discount Microsoft from the potential solutions. Its acquisition of a relatively unknown PPM vendor and heavy investment in virtualization infrastructure make it potentially even more capable of providing an integrated solution that can handle the constant change in resource consumption.

A change in the way projects are assigned to resources and vice-versa in the data center architecture will necessary have a ripple effect throughout IT. From deployment to reporting to IP address management to trending and forecasting and project management, the cloud changes everything.

Follow me on Twitter View Lori's profile on SlideShare friendfeedicon_facebook AddThis Feed Button Bookmark and Share