Cloud Coockoo Land Computing

According to Slashdot, Booz-Allen has analyzed the cost of cloud computing for the federal government in a new report

So what does the report actually say?

As with many articles written by consultants, this one spends a great deal of time introducing Spurious Three Letter Acronyms (STLAs) and almost none explaining or justifying its underlying assumptions.

The short answer is that the government can best realize the cost savings of cloud computing by moving as fast as possible. It is thus regrettable that "there are currently no security standards for cloud computing".

In other words, we have an analysis of the cost/benefits of cloud computing that is forced to admit that the brave new world of cloud computing suffers from at least one significant technical deficiency.

Once this is understood, we can view the reams of jargon in a somewhat more skeptical light. The article is heavy on conclusions but the assumptions leading to those conclusions are hidden in the proprietary Booz-Allen "detailed cost model".

Forgive me for being a skeptic on this, but hasn't every development in computing infrastructure offered lower costs? When was the last time one actually did? Despite predictions, paper consumption increased rather than decreased as a result of the paperless office. It is only recently that electronic displays have become good enough and ubiquitous enough to rival paper. And the largest factor in the current decline in the demand for wood pulp is the displacement of newspaper by the Web.

Cloud computing certainly offers major cost savings in certain specific types of computing environment. But talk of 'the switch to cloud computing' suggests that cloud computing guarantees significant cost savings in every type of computing environment.

At no point in the Booz-Allen article do we learn where these cost savings are to come from. It is implied that some of the savings will come from lower expenditure on hardware and power as a result of better utilization of the underlying resources. There is also what should be a very clear red flag in the assumptions:

"Existing application software will migrate with the infrastructure to the cloud. Application software support costs remain out of scope. "

While some applications will migrate to a cloud environment without issue, those of us who have experienced government computing environments know that even minor changes can require considerable time and days of expensive consulting effort. In a government environment the costs of failure are high. The processes that control change to critical computing resources are designed to mitigate the risk of failure. Computing resources are comparatively cheap compared to consulting manpower. It is by no means obvious that cloud computing will be a break even prospect for the typical government data system, let alone a source of savings.

It is certainly rather difficult to understand how a $3 million investment in cloud computing infrastructure would result in a reduction of 'O&S' expenses from $77.3 million to $22.5. No explanation is given for these figures except for the admitted omission of the costs of migrating applications. While hypothetical cost savings of $50 million for a single data center might appear to be impressive, it represents only a hundred man years of consultant time, an amount that can easily be consumed several times over when an agency is required to make substantial configuration changes to every application running in the data center.

The key oversight of the article is the fact that no distinction is made between adopting the cloud computing model for new infrastructure deployments as opposed to 'switching' existing deployments to the cloud model. This distinction is critical when we look at the costs that the model is focused on:

4.Our model focuses on the costs that a cloud migration will most likely directly affect; i.e., costs for server hardware (and associated support hardware, such as internal routers and switches, rack hardware, cabling, etc.), basic server software (OS software, standard backup management, and security software), associated contractor labor for engineering and planning support during the transition phase, hardware and software maintenance, IT operations labor, and IT power/cooling costs.

These are of course costs that are typically incurred early in the deployment of a specific application. Once a system is deployed these are sunk costs that will not be recovered through a 'switch' to the cloud. While a switch to the cloud may allow a reduction in the cost of power and cooling, this benefit must be set against the real costs of making a significant change in the configuration of a deployed system, costs which are omitted from the Booz-Allen model.

The only scenario in which it is likely that cost savings of the magnitude anticipated in the study might be realized is if the cloud computing model is adopted before the expenditures are made on hardware infrastructure when a new application is deployed or a major revision made to an existing application.

While the savings in such circumstances may well be significant, it is important to describe them as a development choice and not the 'switch' described in the article. Since only a small percentage of government information technology applications are newly deployed or substantially revised in any given year, it follows that the expected savings from cloud computing will also be modest in any given year, if indeed it is possible to reliably measure them at all.

Since the savings from cloud computing are likely to be considerably more modest than those promised, the urgency for action is likewise reduced. Rather than committing to 'switch' to the cloud computing model as quickly as possible, government agencies should only proceed at the rate justified by actual cost savings from actual trials.

Cloud computing certainly offers significant advantages to certain enterprises for certain types of computing services. In particular it is likely to be most relevant for the small enterprises that have no computing staff whatsoever, let alone dedicated 'data centers'. The more often cloud computing is presented as a panacea, a magic wand that automatically delivers dramatic cost savings with little effort, the more practitioners are going to dismiss it as yet another passing fad that promises much and fails to deliver anything.