Following my presentation entitled “Why A Major International Construction Company Needs Cloud” at the recent OpenStack Design Summit in San Antonio, I’ve been asked a lot of follow up questions on two of the points I made during the talk:

1)   Server virtualization “not being particularly attractive” to enterprise CIOs

2)   Why I think it is important that OpenStack itself (via the core “products”) goes someway to differentiating themselves from other available “IaaS” solutions by “moving up the stack” and addressing the real challenges that I see becoming more prevalent in the enterprise space

My trains of thought seems to have caused some consternation (and / or “meh”) in certain quarters, so I thought I would take this opportunity to clarify my comments and offer a little more in the way of context as I think some people were either:

a)    Not physically there during the presentation and have picked up out of context comments and threads via twitter

b)   Asleep, catatonic or heads-down in their laptops during the presentation and missed the end-to-end story, which is based solely on my experiences of the last few years during the build out of  “our” next-gen global IT platform *

* For the purposes of this post, I won’t name the organization I work for. If you were there, you will know it. Google will also tell you easily enough.

Firstly, let me give a little background. Over the last three and a half years, I have been part of a relatively small team that has architected and deployed a brand-new IT environment for a large multinational organization. This has been done partly by consolidation, partly by changes in operating models, partly by ruthless standardization of the back-end platform and partly through new the adoption of new technologies.  However, all of it was done with strong business alignment – through our constant engagement with the business, we knew that the way we were “doing IT” was not in line with the changing needs of the business and hence we had to innovate – and quickly, with their understanding and buy-in. I would go as far as to say that today’s Enterprise IT can’t effect meaningful change without this kind of alignment.

The new environment is what I lovingly call “the accidental cloud”, simply because our efforts around it were planned for and started long before every single conversation around IT included the words “public, private or hybrid”.

The architecture and deployment started with infrastructure, largely because the IT department owned the budget. The infrastructure included new global network(s), new global data centers (reducing the total from “many” to three), and above all, it included virtualization technology, both for the “servers” and the “applications”. To clarify, by “applications”, I am purely talking about the “client” portion of the many, many client / server applications that we operated, and still operate, as LoB today.

You can think of that application virtualization approach as tactical, it’s a “bridge” from the old to the new. The smarter ones will figure out the underlying technology easily enough. It’s not exactly the moon landing.

So that brings me on to server virtualization and as I mentioned during my presentation, we have over 95% of our production servers (somewhere north of 2500 VMs) running as “virtual servers” today, supporting our application and service portfolio.  Overall, we are very happy with our hypervisor and as with any hypervisor, it has brought some obvious efficiency, but here’s the main problem:

We are highly virtualized BUT NOT highly automated.

There is no doubt, nor will I contend, that virtualization, in any format, but specifically for servers, is a key part of cloud, whichever derivative you like…but…servers run applications. Applications automate business processes. Business processes make organizations money and profit. Profit is what most business is about.

Today’s applications, both in our environment, and I would suggest in many other enterprises like ours, are usually made up of more than one server. There are very few (if any) true LoB applications that are single server, single stack (a la LAMP).  They are a combination of n-tiered client / server or n-tiered web applications that have multiple servers stitched together to make that “workload”. This isn’t so much a legacy as a reality – would we build them the same tomorrow as we did yesterday? Of course not. But the harsh reality is that it would cost a prohibitive amount of money to “cloudify” our 800+ “legacy” applications for a perfect cloud world and we can’t find anyone to bankroll that. Could you in your enterprise?

When our line of business people ask for the application to be made available for a new project or department, they ask for just that – the application – not the ability to have IT spin up individual servers and have them handed over such that they can configure their own pieces to get their application functional. They want the application, and almost always, “need it ASAP”.

Now here’s a somewhat generic problem with ASAP.

  • No automation = constant manual intervention.
  • Constant manual intervention = greater time to deploy (x more defects)
  • Greater time to deploy = lag to the business.
  • Lag to the business = potential loss of revenue.
  • Potential loss of revenue = frustrated business leaders.
  • Frustrated business leaders = pressured CIO.
  • Pressured CIO = (well, you know the rest, I’m sure)

In summary, I would strongly contest that the CIO isn’t looking at server virtualization as a panacea, nor does he care too much about the underlying hypervisor technology other than from maybe a price and / or lock-in perspective.

I’ll be willing to bet a few bucks that the CIO is looking at how long it takes to predictably and repeatedly delivery applications to the business (if that means SLA, then so be it). Does the CIO truly get excited about server virtualization alone? You’d be hard pressed for me to say “yes” to that.

Being smarter around how to define, deploy and manage the entire application workload on top of the enabling cloud platform is where I think the wise enterprise architects need to be looking, if they aren’t doing so already.

So, now that clears up my comments around server virtualization, I wanted to clarify my points (and hopes) around OpenStack.

First and foremost, as I have mentioned many times before to some of the key players, I think it is a fantastic opportunity for the OpenStack community to win the race to be recognized as the first true, mature, open and usable platform for both service providers and enterprises who are looking to build or offer cloud services.

By the nature of where the lifecycle of the platform is today, there is some way to go until it is polished, but this is a rough diamond, no doubt about that. The energy around the recent design summit, combined with the commitment of many large tech vendors, suppliers and OEMs is testament to the buzz and I am sure that will grow over the coming months.

As an open platform, there will be “services and support” opportunities (very probably for the likes of Rackspace, Citrix, etc) as well as a rapidly growing eco-system of partners and emerging vendors, keen to offer their niche solutions to “embrace and extend” the OpenStack platform.

I love this eco-system approach, but I strongly believe that if one of these niche areas remains the “application workload opportunity” that I outline above, perhaps OpenStack is missing a vital trick that puts it right up, front and center, against at least more of what, say, vmware can offer today via vApp. The fact is, rightly or wrongly, the typical enterprise is wary of niche vendors.

In my very humble opinion, this is far too important and far too good an opportunity to not be part of the “core” OpenStack thinking.

I’m both excited and nervous, so I will cut to the chase as I round out this post.

Today, I don’t know how I would sell and enterprise CIO on the real value of rushing out and deploying the OpenStack platform. Before I’m shot down, I am fully aware that there are not very many other “cloud” platform vendors who offer the utopia that I am describing, but that is exactly the point – this is a unique opportunity for a open platform (hypervisor and hardware agnostic) to become the central cog in the wheel of enterprise application deployments as they work their way from the past to the future.

I’ve experienced how hard this is without it. All I can do is ask, I suppose.