The Economics of the Cloud: Dissecting a key white paper
- 12 November, 2010 06:25
If you want to understand the key driver of the cloud computing revolution, you owe it to yourself to read Microsoft's new white paper "The Economics of the Cloud." In it, authors Rolf Harms and Michael Yamartino lay out an analysis of the economics that underlie cloud computing, and demonstrate in a convincing fashion why the shift to this new technology platform is inevitable. A copy of the paper is available as a download from the blog posting by the authors, located here.
After a brief introduction, the authors lay out a central thesis: despite initial concerns about shortcomings in new technology offerings, "historically, underlying economics have a much stronger impact on the direction and speed of disruptions, as technological challenges are resolved or overcome through the rapid innovation we've grown accustomed to."
To quote another sage on the same subject, "It's the economy, stupid."
The authors note that this pattern of adoption despite technical concerns is nothing new in the tech industry, citing client/server and virtualization as examples. Their point is that economic advantage ultimately overrides initial reservations, resulting in the spread of the technology.
I would go even further than that. I would say that, given significant economic advantage, organizations will adopt technologies despite their manifest and ongoing shortcomings. If one looks at state of personal computers today, it's scarcely believable that we use a technology so fraught with fragility, security holes, high administrative and maintenance costs, and constant need to upgrade. We put up with those things because of the tremendous benefits we receive by using personal computers, among them easy digital communication, highly modifiable digital documents that facilitate incremental improvement and what-if scenario exploration, scientific problem solving, data analysis, and more.
The question then becomes not are there problems associated with cloud computing? There are, without doubt. Beyond the obvious ones commonly discussed - like security, SLA levels, integration complexity, others like software licensing, bandwidth availability, need for new application architecture design pattern learning present themselves as well.
Instead the question becomes, does cloud computing offer enough economic benefits to cause us to mitigate or live with those shortcoming, a la personal computers. To this question, the authors respond with an unqualified yes.
The basis for the economic advantage is the economy of scale available to cloud computing data centers. The paper identifies three areas of scale advantage:
1. Supply-side savings: Cloud data centers have lower costs per server, based upon purchasing power. 2. Demand-side aggregation: By supporting a mix of tenants, cloud providers are able to achieve higher server utilization rates than is possible for single-tenant data centers (even those for large companies with different departments, business units, or subsidiaries sharing the data center). 3. Multi-tenancy efficiency: Hosting multiple tenants spreads administrative costs and reduces cost per data center user.
One supply-side cost advantage include cost of power, in which cloud providers seek out low-cost geographies and, once a geography is identified, negotiate extremely low rates based on power consumption volume. This article from Computerworld (titled "Data center density hits the wall") illustrates just how critical power management is in today's data centers. Suffice it to say, managing the use and cost of power is a significant input to cost-effective data centers, and cloud providers are experts at this factor. Others include labor efficiency and security (the scale of these data centers allows increased security spend in total, but lower spend per server).
To give an idea of the extent to which the scale is being taken, the white paper lists a number of recent data center buildouts, the largest of which is a 3 million (!!) square foot data center in Scotland.
On the demand side of things, cloud providers support lots of users and smooth out peaks and valleys by aggregating large volumes of compute use. The authors calculate that at very large server volumes, utilization can be increased significantly, thereby reducing operational costs.
Finally, of course, cloud providers tend to run multiple data centers with centralized NOCs, reducing operational costs in terms of headcount. One Microsoft data center, featured in Forbes Magazine, operates a 700K square foot facility with only around 35 people on premise. This is achieved, in part, because the operational personnel responsible for that facility are located off-premise and manage a number of other Microsoft data centers. (Forbes overstates the small number of staff on-site in its article; nevertheless, the ratio of server to employee is remarkable).
In concluding the section on economies of scale, the white paper goes on to state that economies of scale are so significant that the TCO (per server, presumably) of a 100K server data center is 80% lower than that of a 1K server data center.
The paper goes on to explore the implications of this drastically lower cost of data center operation. Among the items it identifies as affected by the new platform are:
1. More use of IT. Even though the entire industry is built upon elasticity of demand facilitated by the reduced costs caused by Moores Law, many people fail to recognize that cloud computing is going to vastly increase the amount of IT being used. Lower cost = increased use.
2. Elasticity of use increases high compute task implementation. When it's as easy to use 1000 servers for an hour as one server for 1000 hours, more compute-intensive tasks will be performed. Think genomics, number crunching, and so on.
3. Reduced risk and more experimentation. When computing does not require capital investment, it's much easier to justify trying something out, rather than analyzing it to death. More IT-enabled business initiatives will result from cloud computing.
The paper then ventures into what is likely to be its most controversial element, evaluating the economics of public versus private cloud computing. Echoing the original UC Berkeley RAD Lab Cloud Computing Report, the authors conclude that, even for a cloud with 1K servers, the private variant runs a 10X higher TCO factor compared to a public cloud computing environment.
The authors then discuss what types of business entities are most likely to adopt private cloud computing in the near-term, and how those choices will transition over time. Their overall position can be summed up as public cloud computing economic advantages will come to be recognized, and concerns regarding security and reliability will gradually fade, thus resulting in a marked preference for public cloud computing.
I feel this is too gentle a depiction of the process. If the cost advantage of public cloud computing is really 10X - and that advantage becomes known and accepted - it's going to be very difficult to justify investing in building out a private cloud environment. If anything, widespread recognition of these cost comparisons would be likely to dramatically accelerate public cloud computing adoption. I recommend the authors of the paper publish their cost analyses to enable others to verify their assumptions and calculations, because, as noted, this section will undoubtedly be the most controversial portion of the entire document.
If you are planning your cloud computing initiative, or getting ready to evaluate your strategy, you owe it to yourself to download and read this paper. It's not often a single document helps shape the entire debate on a topic, but this white paper will come to be seen as a watershed event on the topic of cloud computing economics.
Bernard Golden is CEO of consulting firm HyperStratus, which specializes in virtualization, cloud computing and related issues. He is also the author of "Virtualization for Dummies," the best-selling book on virtualization to date.
Read more about virtualization in CIO's Virtualization Drilldown.