For how surprising this may look like, the  justification is indeed very clear. The Air Force is interested in expanding an earlier pilot where it has been using 300 PS3 to experiment a high performance computing architecture based on the Cell Broadband Engine Architecture, which powers the PS3. This research aims at determining  what software and hardware technologies will be implemented in military systems.

As described in the justification review document attached to the RFP,

With respect to cell processors, a single 1U server configured with two 3.2GHz cell processors can cost up to $8K while two Sony PS3s cost approximately $600. Though a single 3.2 GHz cell processor can deliver over 200 GFLOPS, whereas the Sony PS3 configuration delivers approximately 150 GFLOPS, the approximately tenfold cost difference per GFLOP makes the Sony PS3 the only viable technology for HPC applications.

The document provides further justification about why the Air Force is using PS3 technology, but the one above is quite compelling.

This is a pristine example of how consumer technology can be used in pretty demanding government contexts – although still in an R&D rather than operational capacity. It should make reflect many of those who are still dismissive about the commoditization of government technology (such as the use of consumer social software and cloud computing).

Two main take-aways for me:

• Government clients are confused as to (1) whether vendor offerings really meet their security requirements, (2) which workloads could be easily moved to a public cloud and (3) what’s the roadmap from public to private/community clouds.
• Some vendors have great confidence that their existing solutions already meet most of the security requirements and claim that there is already a lot of government stuff in the cloud.

Therefore, either vendors are not effectively or convincingly marketing their success stories, or government clients use “security” as a blanket topic not to move to the cloud.

My current reading though – and I know this won’t be welcome by some – is that most of these conversations really are about alternative service delivery models.

Except one or two cases, all conversations focused on “cloud computing can make me save money” rather than “I need scalability or elasticity”.

But even just looking at cost savings, most compare their current costs with numbers that come from “public cloud” vendor offerings. On the other hand it is clear that certification and accreditation to comply with a variety of security requirements, as well as additional constraints about service levels (such as finer-grained control of data location) will increase costs. Last but not least, I’ve heard two clients mention the “geopolitics” angle that I recently covered in a blog post as well as in a research note (Gartner clients only).

All this led me to tweet this simple line: In most cases government organizations that mention cloud are just looking for better sourcing, and don’t really need a cloud.

By one of those coincidences that make life so interesting, pretty much at the same time NASA was announcing that the Office of Management and Budget is moving USASpending.gov to NASA Ames Research Center’s cloud computer infrastructure called Nebula.

On Tuesday Oct 20 at 12:30 pm I will moderate a vendor panel with Microsoft, Google, Salesforce.com, Terremark and CSC to have their views about where they think cloud computing in government should and will go and to provoke some discussion about whether they think the will cooperate or rather compete with NASA Ames and the likes.

Interestingly enough, when answering my question, Vivek suggested that providers such as NASA or DISA would provide services for specific government constituencies (such as scientists or defense agencies). However the USASpending.gov, which he did not mention on stage, shows that we are at the brink of interesting times, where federal agencies might have choices between internal or external cloud services. How the GSA cloud storefront Apps.gov will evolve in response to this remains to be seen.

Incidentally, for those who look at internal government cloud solutions as a better option from a security standpoint, another coincidence was an article by NextGov about the vulnerability of systems and data at NASA.

Vendors in the panel (in strict alphabetical order) are:

• CSC
• Google
• Microsoft
• Salesforce.com
• Terremark

Their choice is neither reflective of their position in the market nor of any assessment about their capabilities and strengths with respect to other vendors. All we wanted to have was a sample vendors who are currently discussing with their clients and internally about opportunities and challenges of cloud computing in government and the broader public sector.

As this is quite an interesting line-up, I thought I would ask readers of this blog to suggest questions they’d like to ask. I am already getting a few on the Gartner Symposium – Government 2.0 group on LinkedIn.

So, don’t be shy and shoot your questions by just replying to this post. I promise I’ll pick the most interesting ones for the panel and will come back with the outcome of the panel both here and (for Gartner clients) with a fully-fledged research note.

Now RACE is available also as an operational environment, according to what Henry Sienkiewicz, technical program director for DISA computing services, said at press briefing, as reported by NextGov.. The service will be first available available on unclassified defense networks, but is expected to be extended to its secret network soon.

This is an interesting example of how some federal agencies are positioning themselves as providers of private or community cloud services. As I said in a previous post, it will be interesting to watch what their role will be in the evolution of OMB-GSA strategy on cloud computing (whose first step has been Apps.gov) and whether they will find themselves competing with vendors who move into the private cloud business.

Henry Sienkiewicz will join our panel on The Government and Cloud Computing to be held on 18 October on the first day of Gartner Symposium in Orlando, immediately before Vivek Kundra’s interview. The panel will be moderated by my colleague Jeff Vining and other panelists will be Casey Coleman (CIO at GSA), Chris Kemp  (CIO at NASA Ames Research Center), June Hartley (Associate IT Director at NBC), and Gopal Khanna (President NASCIO and CIO Minnesota).

Besides that, yesterday I had an exchange with a few Gartner colleagues about the thorny issue of data location. Vendors like Google claim they will be able to provide government clients in the US cloud services that comply with federal security requirements (FISMA) and run on servers that are physically located in the US. Our discussion was whether this will be enough also for state & local as well as international government clients to be moving their email or desktop applications into “the cloud”. Of course this does not apply only to Google but to any vendor providing cloud services of sort.

None of us is a legal expert, although a few have a legal background, nor is Gartner supposed to provide any sort of legal advice. However there a number of issues that apparent in discussion with clients. For instance:

• what if, during an investigation in a particular jurisdiction (say, a state), law enforcement authorities need to seize data concerning a subject who is being investigated? Could they seize data located in a different state (as opposed to data on a local desktop)?
• what if a server seized for an investigation in a particular jurisdiction contains data concerning subjects who reside in a different jurisdiction? Would this infringe their data protection rights?

Should this not be an issue within the US (I frankly do not know), it certainly is across different countries.This, combined with the desire of showing a positive impact of government IT spending on local economies, creates an interesting set of issues for cloud service providers.

The question is whether cloud computing can still deliver its potential benefits if clients are granted data location control at a pretty granular level. Of course this is not a problem if clients are federal government agencies and the jurisdiction is the whole US. But if you think about Europe, with its many countries, some of them being in turn federal, things get far more complicated.

This is particularly relevant for private and community cloud services. The large public cloud vendors, such as Google, Amazon and Microsoft, will compete with infrastructure utility providers, such as HP or IBM (that already run infrastructure and application services for several government clients) as well as with government-owned shared service centers aiming to become cloud service providers for their jurisdictions. Whereas the scale of large global-class infrastructures from the former should provide a better price-performance ratio, procurement decisions will be influenced by a variety of other criteria, including data location control and public value impact.

Definitely the jury is still out on who will win this race.

But the highlight of the event will be a keynote interview with Vivek Kundra, the US Federal CIO, that my colleague Tina Nunno and myself will be conducting . This will take place on Sunday October 18 at 6 p.m. and will last about 45 minutes.

As Vivek is a Web 2.0 enthusiast, I thought it would make perfect sense to crowdsource some of the questions we are going to ask him.

Therefore you are welcome to suggest your questions in four different ways:

• by posting a comment to this blog
• by posting a question on our Gartner survey page
• by posting (or rating) a question through Google Moderator  (it requires a Gmail account)
• by tweeting with hashtag #GartnerKundra

I will be issuing a couple of reminders and possibly a summary of top questions as we get closer to the event.

Looking forward to being joined by some of you on stage and by most of you in the audience.

…the interrelationships between government and cloud service providers will always go well beyond a simple client-provider one. Contentious points in deciding for or against cloud-based solutions will necessarily include elements such as local economic development (do we attract businesses in our jurisdiction? do we create better opportunities to exploit broadband investments?) and employment both inside and outside government (do we put government jobs at risk if we outsource to cloud-based solutions? do we create opportunities for local businesses to leverage from this?).

Yesterday I had all the above confirmed in a revealing conversation with a director in a shared service organization serving a European province.

They are in the process of re-insourcing infrastructure management after a troubled outsourcing experience. I challenged him saying that this may seem a backward move and mentioning all the conversations I am having with government clients who are looking for ways to get rid of managing infrastructure and for alternative sourcing models, including cloud computing.

He reminded me that for a provincial government it is quite important to ensure that IT service or product contracts have at least in part a tangible return for companies in the province. So, while he realized the potential of cloud computing, he said that it would hardly allow them to control how their spending would benefit local players. So, even if cloud service providers were running servers in the province, and assuming they would offer private cloud services that keep the government’s workloads within the province, they should still be able to demonstrate an impact in terms of purchasing products or services locally.

So, forget remote server management, centrally procured equipment or telecom services and an offshore helpdesk; the cloud service provider may have to deal with enough constraints to make the whole model either unfeasible or too expensive to be a valuable alternative.

This made me reflect because, once cloud service providers look outside of the US federal market or some large national jurisdictions (in non federal countries), they will hit these problems. They may not be formulated as clearly and explicitly as our client did, but would manifest themselves through procurement constraints or preferences of sort, or would simply lead potential cloud service clients not to consider the option.

The links between best value procurement and industrial policy are quite complex and need to be factored in by all those who believe that cloud computing is such an appealing proposition that government clients simply “can’t refuse it”. Until vendors will be able to spread the benefits of their cloud services to the jurisdictions they serve, they’ll have a hard time selling into them. Compliance with security, availability, e-discovery and other requirements may prove not to be enough.

Shortly after the official launch of the GSA cloud storefront Apps.gov, vendors that feature multiple times in the business and productivity application sections of the web site have leveraged their position.

Salesforce.com issued a press release  about its inclusion in Apps.gov, while Google announced Google Public Sector, a directory of existing Google offerings packaged with information that speak to a public sector audience. As reported by Infoworld yesterday, in conjunction with the Apps.gov launch, Google also announced the availability of cloud services designed specifically for US government agencies earlier next year (by when it aims at getting the required FISMA certification) – see also Google’s announcement on their blog.

As a side issue, I tried the new PublicSector page in other countries too. In the UK and in Australia, the page is pretty much the same as in the US, whereas elsewhere- such as Italy, France, Germany, Denmark, Canada - it is a different version (and not in the local language), with a few broken links on the lefthand side. I wonder whether this is a reflection of how Google sees the government market maturity for its offering,

I expect players like Microsoft, Oracle, SAP and others to move rapidly in this space to establish their role as cloud application service providers. How ready federal clients are to embrace these new models for mission-critical applications remains to be determined.

Also, looking at the current plans of the federal administration, I cannot find much about platform as a service. This is key to address application integration in a scenario that becomes more rather than less complicated, with a combination of public and private cloud infrastructure, SaaS and on-premise applications. I am looking forward to seeing where IBM, Microsoft, Tibco and other aPaaS solutions fit into the US feds’ scheme of things.

The launch has created a big buzz on the Internet, with hundreds of tweets, blog posts, articles mostly supportive of this new initiative. The storefront provides access to four categories of services: business applications (including CRM, ERP), productivity applications (such as collaboration, office suites), infrastructure services (such as storage, virtual machines, web hosting) and social software.

This is part of a three-leg strategy, including also budgetary focus (to support a number of pilots on cloud computing for collaboration and lightweight workflows in 2010 and to provide guidance to agencies to adopt cloud computing in 2011) and policy planning & architecture (addressing centralized certification, security, privacy, etc.)

The social software  section is of particular interest, as it finally reveals to the public the amended terms of use agreed by GSA with some of the social software providers (see here for the Facebook example).

I do certainly join other people who congratulate GSA and OMB for this achievement. However I’d like to offer some initial observations about areas for improvement and clarification:

• All the available applications are for the so-called “public cloud”, so they can be used where agencies have no need to control where data are being located. I do not know yet (and I am trying to ascertain) whether licenses for products currently on the store provide any guarantee about data location. I assume this was part of how those applications were selected, but it is not clearly explained in the site, nor in its (relatively succinct) FAQ section.
• It is not clear whether prices are monthly, per user or what. It would be better to make it clear upfront without requiring users to find it out following the links to different products.
• Vendors like salesforce.com and Google are predominant. While this probably reflects market maturity and is meant as a stimulus to other vendors to catch up, it looks a bit weird from a public procurement perspective, as some may read this as an endorsement of those two vendors. However in his address, Kundra mentioned other vendors (such as Microsoft or Adobe) that will soon join the storefront.
• The infrastructure services section is still empty, as there is a GSA RFQ out, which concerns the procurement of public cloud infrastructure services. It may be worthwhile providing a pointer to the RFQ from apps.gov itself, so that agencies start familiarize with different categories of services.

It is quite clear that so far Apps.gov addresses exclusively the “public cloud model”. It will clearly play a crucial role in exposing agencies to the use of consumer social software, and provides a value added for those who have already articulated their own SaaS strategy. Minor improvements on the FAQ section, to pricing information and in giving prominence to how the terms of use address federak requirements would certainly help clarity and transparency.

The real battlefield, though, is going to be around the so-called “private” and “community” cloud models. It is interesting that the launch took place at the NASA Ames Research Center, which is also known for its own cloud infrastructure initiative (code-named Nebula). Vivek was very complimentary of what NASA Ames has achieved and highlighted it will have its role in the government “private” or “community” cloud infrastructure. However it is a fact that moving from acquiring new apps or external computing capacity that meet new requirements, to rationalizing existing workloads and balancing them across previously siloed infrastructures (currently managed through a variety of sourcing models) is a totally different challenge.

Budgetary and policy planning measures on security, privacy and procurement, will definitely help, but it is the enterprise-wide governance bit where most shared service initiatives have failed or struggled in the past. Why should the federal cloud initiative be any different?