Blog post

Is Google the Mainframe of Cloud Computing?

By Tom Bittman | September 26, 2008 | 4 Comments


I’m currently flying above 30,000 feet, well above a solid bank of confused clouds. Seems appropriate to think about the structure of cloud computing in the future.

Conceptually, cloud computing services can and will be provided at any layer of the IT computing stack, from raw compute services to business process services. So, conceptually, a cloud service might actually include a supply chain of cloud services – a cloud process provider using cloud software providers using cloud compute and storage providers, etc.

But that isn’t what’s happening today.

In these pioneer days, cloud service providers are doing it all. Software as a service providers are building data centers filled with customized infrastructure – and in Google’s case, highly customized and proprietary hardware. They are building, essentially, silos of vertically integrated solutions in the cloud. To be sure, these are cloud services, but why aren’t they using cloud services themselves?

One answer is that it is early, the supply chain, as such, doesn’t exist, and what does exist is not standardized. Another answer is that many cloud services are highly-focused and purpose-built (e.g., web search), and therefore underlying cloud enablers do not need to be general-purpose, and in fact, are more efficient if they are customized to do just a few things well.

But what happens when the supply chain starts to emerge? When standards improve for accessing services? What happens when the cloud service provider wants to expand into new businesses?

I believe that the future of cloud services will be much more dynamic and fluid, with both horizontal and vertical federation. While highly integrated and proprietary vertical cloud services will exist, they will be a minority – and in fact, their competition will be fluid, federated solutions of smaller competitors who should be much more agile. A future challenge for existing vertically-integrated cloud service providers will be to componentize their offerings – open up cloud services at lower levels of their cloud computing stack, and leverage other cloud service providers when appropriate. To be sure, they can differentiate by having customized hardware or unique management software, or unique performance capabilities – but betting on this differentiation going forward seems extremely risky as the markets mature.

This is exactly what happened to servers – they went from highly integrated and proprietary mainframes (proprietary hardware and limited applications) to midrange systems (with packaged software from third parties) to high-volume servers (hardware separated from operating systems separated from business applications). Mainframes have not gone away – but their role has certainly become a market niche. Mainframes have been attempting to evolve for years – to support more operating systems, more packaged applications, etc. This has slowed but not stopped a declining role for the mainframe in global computing.

Will Google and become the mainframes of the cloud? Will early market leadership translate into strategic viability even in a more componentized cloud services market? Will they invest strategically to avoid a vertically-integrated gridlock? Should Microsoft be building datacenters, or instead, directing datacenter standards for Microsoft software? Hmm…

The Gartner Blog Network provides an opportunity for Gartner analysts to test ideas and move research forward. Because the content posted by Gartner analysts on this site does not undergo our standard editorial review, all comments or opinions expressed hereunder are those of the individual contributors and do not represent the views of Gartner, Inc. or its management.

Leave a Comment


  • Interesting thoughts. On your concluding questions, Well it turns out we (Microsoft) are doing both – partly because we believe we can do a better job of providing specialized orchestrations of Software + Services on a grand scale from datacenters of our own design and specs – because the datacenter is in some ways becoming a grid OS itself, and we think we know something about how to do OS’s. In fact, we’re testing that out on Amazon’s EC2 right now (I wrote about it yesterday at with Windows Server there… not sure why they were kind enough to let us “beta” our own plans on their platform but hey we’re cool with it! So we’re going to wind up doing both routes you mention.

  • Niraj J says:

    Tom , your post kinda assumes that googles wants to be a provider (Seller) in the cloud computing space(defined by one of the layers in SaaS , PaaS , HaaS) essentially an arms supplier to businessness.

    Google is really a using the cloud computing paradigm as an enabler to essentially own the battlefield where businesses fight as compared to being an arms supplier. It will go bottom up – and focus on being the IT platform for small business as a whole to run their business.

    refer link here for more details

  • alekya says:

    I had a doubt as google needs store lage amounts of information will it use mainframes if so when did it started using it

  • alekya says:

    I had a doubt as google needs store large amounts of information will it use mainframes if so when did it start using it from initial stage itself???????