Blog post

Building a Private Cloud: Are We There Yet?

By Tom Bittman | February 17, 2009 | 4 Comments

Future of InfrastructureCloud

My last blog entry caused quite a ruckus. While the term “private cloud” has been in use for at least a year, there are still semantic, moral and religious issues with the term. I respect the issues, but I have also found the anti-private-cloud zealots to be a minority. In the past week alone, I met with fifteen different companies (vendors and users), and most of the conversations revolved around the private cloud concept. All of the vendors are embracing the term (no surprise). More importantly, the users were also embracing the term, the concept, and the strategy.

So how do you know you’ve arrived? What is different from a private cloud versus a well-run IT organization?

I can think of four fundamental attributes: measurepriv

  1. Services-oriented. The IT organization sells services, and the IT customer only deals with IT through service offerings (self-service).
  2. Variable Pricing. Based on usage. The IT organization maximizes technology sharing and automation for efficiency.
  3. Elasticity. Customers deploy/grow/shrink/retire services quickly. IT is architected to ensure fluid scaling on demand.
  4. Independence. The IT organization could move a service to the public cloud by flipping a switch. Likewise, the IT organization could choose to open their services to external customers easily. I even talked to a company today interested in take their services “public” sometime in the future.

There are probably many more, but I think this captures the essence. What am I missing?

Leave a Comment


  • Krish says:

    The actual problem lies in your (and other private cloud evangelists’) definition of service. It is done for convenience to suit the agenda of enterprise users. I don’t believe in the idea of private clouds because in my view, the cloud computing should include the utility nature of computing like electricity distribution. Your definition of service doesn’t include the cost advantage offered by the actual cloud computing definition of service. In the actual sense, the idea of private clouds doesn’t make any sense. But I am not religious about the usage of the term. If “private clouds” is the path which enterprises will take to embrace “public clouds” eventually, so be it. Let us use the term to keep enterprise community subscribed to the concept of Cloud Computing. With further technological evolution and building up of trust in terms of security and common sense regulations, the enterprises are bound to adapt “public clouds”. Those who resist will eventually lose out. Till that day comes, let the usage of the term “private clouds” be in vogue. Let us not fight over these trivial semantics. Instead, let us debate on issues that could move this field further. My 2 cents.

  • Tom Bittman says:

    Valid opinion, but I think the issue is an assumption that a public cloud provider will always have better economies of scale than a “private” one. Even if we assume that all other “service” concerns are equal between a public provider and an internal provider (which will probably become more true over time, but has a way to go), it is not necessarily true that cloud computing services will always trump internal on scale. Here’s a simple example – let’s say the federal government decides that they want to more effectively spend their $40B+ in infrastructure spending every year. They could create private cloud services that are shared by all federal departments. In fact, the government could even sell excess capacity eventually and become a public cloud provider. Could they, instead, move their services to a public cloud provider? One way might be to eventually spin off (non-security-critical) portions of the private cloud to be a public entity. But could they switch to public cloud providers? How many of them, and at what profit margin? Who is bigger and could support all of their services?

    OK, this is extreme, but my point is that there is a continuum here of scale, elasticity, service offerings – and there is a reality that enterprises have lots of stuff already – and drawing the line from here to there will often lead them through “private cloud” over the next 5+ years. And even then, it seems to me we should not assume the path is a one-way street leading to public services only – more likely an adjusting hybrid model. Until computing becomes a fixed commodity and provides zero business differentiation.

  • Sam Johnston says:

    Does the federal government still generate their own power? They did at some point but the grid took over virtually overnight and an entire industry around selling, servicing and fueling generators vanished.

    The same will happen with cloud computing (probably quicker than your average person might think) and then we’ll get back to calling it “computing” again, only it will look vastly different just as it did with the advent of “personal computing”.

    In answer to your question, we don’t have variable pricing in The Cloud (we arguably should and probably will eventually) and it’s pretty optional. CapEx moves to OpEx (and is greatly reduced in the process) so the same should be true here, at least for the “end users”. Not having to deploy infrastructure, human resources, etc. and having a very low barrier to entry (e.g. web based provisioning console) is important here.

    Delivery via web and/or web services is important in terms of device and location independence (really tcp/443 should be the only open port but much of what people are branding “private cloud” is just an evolution of virtualisation).

    Performance, reliability and scalability could probably fall under the “elasticity” moniker. Performance should be monitored and consistent. Reliability should be high[er] via redundant systems and sites. Scalability should be seamless and seemingly infinite (e.g. on-demand requests should be satisfied in almost real-time).

    There’s probably others but in any case “private cloud” PoCs/pilots tend to have one or more public components (even if just something like messaging & web security and/or email) so the term “hybrid cloud” is usually more suitable – it’s what I tend to use to (usually successfully) sidestep the whole debate. Rather than tell clients “private cloud is just like public cloud, only better!” I’m fairly candid about the pros and cons of each.


  • Tom Bittman says:

    Sam, appreciate your comments. Please comment back to me.

    In my opinion, the electricity distribution analogy is overused and doesn’t fit really well. First, we are talking about the future, because the vast majority of enterprise needs cannot be served through the cloud – yet. Second, the primary reason that enterprises are cautious about cloud computing is security of their data. Not really an issue that kept them from using distributed AC power. Also, unlike electricity distribution, IT is evolving at an extremely rapid rate. Finally, the number of enterprises generating their own electricity in 1887 was miniscule compared to the number of enterprises generating their own “IT” today. Even so, the grid did not take over virtually overnight – it took 20 years before utilities produced 40% of the electrical production in the U.S. (1887-1907).

    The innovation in electricity was distribution and standardization of alternating current, replacing DC (which couldn’t transfer over long distances). It wasn’t an economies of scale innovation. In many ways, the invention of AC was like the invention of the Internet. So why hasn’t computing across the Internet replaced enterprise computing yet? In IT, the “distribution” mechanism has been in existence for more than a decade. What’s really changed are technologies that enable economies of scale and sharing – multitenant apps, virtualization in many forms, heavy automation of standardized architectures. We should not ignore the fact that enterprises can gain some economies of scale themselves, internally, for example, through virtualization. A 12 to 1 consolidation/pooling ratio of virtual machines, for example, is not an economies of scale technology you can ignore – and enterprises aren’t ignoring it.

    I agree variable pricing is optional. The shift to OpEx is the most important financial consideration – and the more the OpEx is related to usage, the better. That could include subscription pricing, tiered pricing, or variable usage pricing – but it should NOT be related to physical implementation costs – those are shielded from the user.

    I agree with you that the term “private cloud” will be abused, and will be used by some as a delay or defense mechanism against cloud computing. I want to promote it the opposite way. Private cloud computing principles, defined correctly, should be the most pragmatic path for most large enterprises toward public cloud computing. While they are waiting for public cloud computing services to mature, there is much an enterprise can do to prepare to leverage those services in the future, and gain some benefits internally.

    At the same time, we are telling enterprises to experiment now, and move to the right public cloud services now, and build the decision mechanisms internally to make these sourcing decisions, now.

    I am a believer in the concepts and benefits of cloud computing. However, I also believe that enterprises are not going to make a big switch virtually overnight – they will make a gradual shift. I just want them to do it proactively, and to start now. Why wait?