REALITY CHECK—Commentary

Why cloud computing is still a red herring

My last column, “6 trends government IT managers should be wary of” received a good amount of praise and criticism. The key point of the article was that government information technology requirements — in terms of records and data management, security, and reliability — are very different from business IT requirements. At the most basic level, government is not profit-driven, and its needs for accountability and transparency require a strategic emphasis on IT.

One criticism was that I did not offer enough evidence, and I hope to rectify that by following up on each technology individually (I also discussed the article in an interview on Federal News Radio). Several critics also accused me of being archaic and a defender of the status quo, but nothing could be further from the truth. Throughout my career, I have implemented and pushed advanced technology as a strategic advantage, and I will continue to do so. The article’s critique of each technology exposed problems, not necessarily fatal flaws; each could still improve and be applied to their fullest.

The accusation that I am some kind of luddite was loudest for my critique of cloud computing. Although cloud computing is promising, the bottom line is that it is not yet finished. Using a cake analogy, cloud computing isn't fully baked. Let’s look at some of its pros and cons.

The pros:

  1. The basic idea and value proposition of cloud computing is compelling. The concept is that computing services of various types — the National Institute of Standards and Technology has defined three: infrastructure, platform and software — can be dynamically provisioned and scaled as needed. Some of that is an evolution of previous concepts, such as utility computing and application service providers.
  2. There are several successful vendor implementations of cloud services, such as Amazon’s Web Service stack, Yahoo’s Build your Own Search Service, Google’s App Engine and Salesforce.com. Unfortunately, those early successes are proprietary and therefore also appear on our list of cons.
  3. Finally, things with strong, simple value propositions — such as the Web, virtual machines and the Extensible Markup Language — evolve and mature until they achieve mainstream adoption. As Robert Heinlein suggests, this usually happened in their third iteration.

The cons (in general and with the General Services Administration’s cloud computing request for quotations):

  1. In general, the success stories in cloud computing are all proprietary (see No. 2 above), with proprietary application programming interfaces. So whose cloud do you want to be part of? Rushing to replace stovepipe systems with stovepipe clouds is not revolutionary progress.
  2. On average, hardware and software costs are small relative to software development costs. That is one reason offshoring has grown exponentially. As an example, in one data warehousing project, the hardware and software costs were 25 percent of the overall cost. This would suggest that infrastructure as a service (IAAS) is not where you get the real bang for the buck; platform as a service and software as a service are.
  3. The GSA Cloud Computing RFQ specifies cloud storage services that are nonrelational. But the federal government has a huge investment in relational technologies, entity-relation modeling and database techniques. So in our haste, are we throwing out years of experience and requiring massive retraining?
  4. In GSA’s RFQ, IAAS data management is inadequate. There is no support for reference data, data asset metadata, role-based security, record-level access, records management, data validation, lineage or cleansing. In my opinion, this is a deal-breaker for today’s incarnations of cloud computing. Some others share this opinion.

So as I said in the six trends article, the best strategy for government right now with regard to cloud computing is to exercise prudence and hold off on major deployments until key issues have been resolved. In the meantime, the government can do three things: run test projects, foster standards and develop a concept-of-operations document. The Web is already revolutionizing government, so we can afford to let cloud computing bake a bit more.

Reader Comments

Tue, Sep 22, 2009

Bravo! This honest evaluation is needed in the Federal space. It is easy to get carried away with charismatic "leadership" and lose site of bringing a basic value proposition to a solution. Regarding google's data liberation.. sure Google will let you have your information at any time you want, after they have reviewed it, analyzed it and data mined it. Here is the real concern. Any security pro will tell you that confidentiality, integrity and availability are crucial for a system to maintain. While availability and integrity seem to be pretty good on the cloud the confidentiality and reporting is worrisome.

Mon, Sep 21, 2009 Paul

Before accepting what Google says thier product can do, how about someone who has done it, a third party experience? Mike is right in many areas of this topic, not the least of which is the need for standards. That said, being able to export content is not the same as having open standards with well known APIs. We have needs the far exceed the GSA RFQ, so until those are addressed I can't see too many agencies using GSA cloud computing except as an experiment. If you need more data about the proprietary aspect of the current cloud computing options visit the sites. If you can find two that use the same API calls then Mike is wrong. I don't think you will.

Thu, Sep 17, 2009 Washington, DC

Mike should be commended for bringing up these issues and having an online debate so that readers can better educate themselves. However, one criticism Mike acknowledged is that he did not offer enough evidence to support some of his opinions. When he says "Unfortunately, those early successes are proprietary," he needs to give readers better concrete examples and more analysis. It seems some companies are moving much faster on this front than others to really addressing the issue so users have complete control and choice. For instance, Google's Data Liberation Front http://www.guardian.co.uk/technology/2009/sep/09/google-data-liberation-export Excerpts "Brian Fitzpatrick is leading Google's drive to make the export of data a one-click process from Google's servers onto a storage format of your choice – whether that's your own web server, your computer, or the comfort of your backup drive that you keep locked away in a fireproof cupboard after using it every night." "Users have never been locked into Google search," Fitzpatrick told eWEEK Sept. 15. "It's very easy to switch, you just click somewhere else. That's served us very well and we think that doing that with the rest of our products will serve us well." "Google has been gradually making it possible for Google Apps business users to move data in and out of Google Calendar and Google Docs, with plans to 'liberate' Google Sites and enable users to do batch exports of files from Google Docs." "Specifically, Fitzpatrick and his team are working to let business users export their entire Sites wikis as HTML with microformats. Users would be able to take and drop the wiki into an Apache Web server." "For Docs, Google is working to let users select multiple Docs files and export them in OpenOffice, HTML or Microsoft formats. Google's servers convert these files into .zip files, and Apps users can e-mail these payloads to colleagues for collaboration." http://www.eweek.com/c/a/Messaging-and-Collaboration/How-Googles-Data-Liberation-Front-Can-Boost-Google-Apps-858940/

Wed, Sep 16, 2009

You really should have talked to Obama's CIO and CTO before they threw the weight of the federal government behind Cloud Computing and SAAS on Sept 15th, 4 days after this "insightful" article was published. Bad timing man!

Please post your comments here. Comments are moderated, so they may not appear immediately after submitting. We will not post comments that we consider abusive or off-topic.

Please type the letters/numbers you see above