Inchoate Curmudgeon

You might not think these things are related but they are.

I was around, and actually participated in, the bloody revolution that left glass-house mainframe data center operations in virtual flames. I was working on large, and I mean to say scary, scary large, construction projects. These projects had a typical burn rate in the hundreds of thousands of dollars a day. The last one I worked on had a burn rate of a million dollars a day, with 10,000 people working 24 hours a day, seven days a week.

I worked on systems, primarily PC based, that helped the project managers and engineering staff get near real-time visibility on where work was done, where it was effective and where it was not. It was done with hundreds of PC’s scattered and networked around the job sites where I worked. The largest system we ever used was a clustered VAX running Oracle 6, that spewed carton after carton of paper, packed on a pickup truck nightly, and distributed to teeming masses of anxious superintendents, craftsmen and engineers. We had a war room just to deal with the nightly busted jobs at 6:00AM, irate phone calls and request after request for more and more information better and faster.

The inability of the corporate IT systems, people and processes to deal with this crushing need for information was almost comical. What would have taken them months to accomplish and get wrong we had to get right, every day, in a matter of hours. I can assure you at the time, none of the corporate people we dealt with ever thought the whole PC and small computer thing was “real”.

So, when I received an email today wherein someone was asking how “real” Dynamic Infrastructure (DI) is, I was a bit vexed. This is a troublesome question, not because its hard to answer, but because its the wrong question.

I could only answer it is all completely in line with our continuing march to a fully automated data center, or rather, our march to make the data center invisible and deliver services not servers. No one is closer to getting this right as a product and service offering than Sun Microsystems, and no other executive management team I know of even has an inkling of a vision of what this is really about.

This perspective has some very difficult and painful angles for Sun, and for our customers and partners. It is no secret that we make our bread and butter selling servers. Our customers are largely data center people. But the painful truth is that they largely don’t understand “services” and don’t know who their customers really are.

But the end users of the value provided by Sun’s products are increasingly frustrated with the complexity, time and cost associated with the whole “data center” proposition. Since they write the checks, we should all pay close attention.

The story of the “IBM PC” that killed the mainframe has a ominous corollary in what’s happening today. Many small organizations are not buying hardware at all, other than their Mac’s, and are deploying solutions directly to full service virtual hosting providers, like Joyent. The Joyent model is moving rapidly to the modern equivalent of the “IBM PC” that killed the mainframe. If we aren’t part of this sort of solution, we will suffer. Worse than that, like the corporate folks I dealt with in my construction days, we will be part of the problem.

Eventually, I assure you, large enterprise customer organizations will begin abandoning their internal IT en masse for cheap providers outside the corporate structure. I’m not talking about the large “manage our whole data center” kind of outsourcing propositions. I’m talking about the by the by-the-sip, host my little corporate application kinds of propositions.

There are thousands and thousands of these little applications. They are expensive. They consume thousands upon thousands of servers and thousands and thousands of kilowatt hours on thousands of data centers around the world. They are on embarrassingly underutilized servers. They take too long to deploy. They are too hard to manage.

By contrast, I just got my brand new Joyent virtual machine, on the Internet, with a real IP, root access and full service stack with Apache, Mysql, Ruby et al in 24 hours. Its in a Solaris Zone running on Open Solaris. If I’m not consuming the cycles, someone else is using them. It’s efficient. It’s green. It’s fast. It’s highly available. It’s cheap. Who needs corporate IT? I’m ready to deploy. Lead, follow or get out of the way.

The only way to longer term survival is for us all in this business to learn how to quickly carve full, highly available, high-performance, end-to-end services from gobs of infrastructure. This is what red shift is about. This is what DI is all about. This is what SOA and SAAS and all of this is about.

I find the fact that someone got a DI briefing and is asking if it is real somewhat sadly amusing. The part that isn’t real yet is the part where mainline IT people wake up and realize their customers are leaving.

Trackback URI | Comments RSS