Man, where to start with this one. Let’s do a personal level-set first. I freely admit, I’ve never been a software guy. I can script, I can hack, I can handle web design with simplistic data-driven back-ends. I can write SQL queries like a champ. I’ve been working in and around VMware for half a decade. But I’m not a software guy. I am an infrastructure guy, and everything that I have done is in support of that infrastructure. I haven’t written any significant code since college, and even then it was more for fun than anything else.
That said, the world is changing. I’ve been part of the movement to make infrastructure boring, and having started to see that come to pass, the interesting stuff is increasingly upstream. Of course the cost of entry upstream is that it’s inhabited by the folks who have been toiling away on the software. The left hand is finally getting to meet the right hand, and the results are…interesting.
I am a regular at infrastructure conferences. EMC World, VMworld, and Cisco Live are the core conferences of my profession, along with their Gartner, Interop and GigaOM counterparts. I have gone to software conferences, but they have typically been the large, enterprise software vendors like SAP and CA, who have almost become part of the infrastructure at this point. OpenStack Summit 2013 in Portland was my first “real” software conference, where the majority of the people attending actually were involved with the creation of the software in question. And I’m here to report that it wasn’t at all what I expected. Not even close. Here’s what I learned:
- The conference was for real. I expected more or less what I’d heard about San Diego. A few people, some super-technical sessions that droned on about brokers and debated the merits of Linux distributions. What I found was close to (or slightly exceeding) 3,000 people, a series of well thought out sessions, a more-or-less consistent theme to the show and a venue that was filled to capacity. In fact, there was more bitching about the fact that it was damn near impossible to go to back-to-back sessions than there was about the quality of the sessions themselves. Of course, I found the trick:
I found the secret to getting a seat for in the popular sessions at #OpenStack: show up an hour early and enjoy the lightning talks first!
— Jeramiah Dooley (@jdooley_clt) April 17, 2013
That doesn’t mean the conference doesn’t have a ways to go. There were a significant number of “pay-to-play” sessions, and the general organization around the keynotes was pretty simplistic. They also need to figure out how better to accommodate the fact that half of the show is people with their heads down coding, with the other half which is the vendor showroom, the sessions and the networking. Definitely not a bad start though.
- The community is extraordinarily well organized. Amazingly so, considering the mix of people and priorities. A lot of credit here goes to Rackspace and the Foundation, who have demanded quality and coordination from early on. The way the work is broken into teams, the way each team has a representative, the process by which code is committed, all of these things were a surprise to me. It’s not a bunch of coders with GitHub accounts. Not by a long shot.
- OpenStack itself, with the Grizzly release, is not perfect, but it is functional in the right environment. Make no mistake, putting and keeping OpenStack in production isn’t a trivial task today, even with a limited number of workloads. The cost of the implementation and support of these environments is still too high, and while the integration vendors have started shaking themselves out, there’s a sense that it’s all “consulting value” and not any real differentiation in the product itself. Complicating this is that too much customization breaks the core of the platform and creates interoperability issues. The ecosystem has to find out where they can monetize the interaction, and how they can compete for customers.
- The vendors have arrived. As expected, the big players were here this week, in force. HP, IBM, Dell, EMC, Cisco, VMware, Red Hat, Microsoft and more all played big roles in the event, which was good on one hand and difficult on the other. The presence of those companies will start the process of overcoming the hesitance of the enterprise customers, but it clearly frustrated some of the developers in attendance that there were things being done that weren’t being shared with the community. The companies who can bridge that trust gap (Red Hat and VMware seem to have done well this week) will be the vendors that the community gravitates to.
- Like it or not, VMware is the key to pushing OpenStack into the Enterprise. Yes, the majority of deployments today are using KVM. The challenge is that it’s difficult to make a case for KVM to an enterprise that has already invested in VMware, and with 100% of the Fortune 1000 using VMware for some workloads that’s a huge number of companies that will struggle to adopt OpenStack in its current form. For them, there’s no evolutionary way to move forward. Yet. It was fascinating to speak with members of the OpenStack Foundation and hear how much value they see in VMware being involved. Today, they have done a wonderful job spreading the word of OpenStack to the “anti-enterprise”: the early dev/ops adopters, the CD/CI development teams, the web-scale companies, the AWS users, etc… While we are starting to see, and the Foundation went to great pains to point them out, enterprise adopters, even those use cases were limited. Best Buy wasn’t moving to OpenStack for their IT environment, they are moving their web application. Samsung wasn’t moving their messaging, payroll, HR and finance apps to OpenStack, they were moving their scale out development apps. For OpenStack to become more than the defacto platform for the “anti-enterprise”, they need to be able to piggyback on VMware. Let’s face it, VMware has gone through this process with the enterprise before, offering an evolutionary “path to the cloud” specifically as the alternative to rewriting applications and changing cost models. The “Three Paths to the Cloud” and “Journey to the Cloud” are narratives that VMware have used to help customers evolve from the data center owning, hardware buying enterprises to the private cloud/hybrid cloud customers of today. I know that no one likes to hear it, but VMware is the key. It won’t be easy for them, because in order to embrace this framework there’s some significant changes they will have to make in their revenue streams, but it has to happen if OpenStack wants to progress quickly.
- Rackspace has pulled off a miracle. I’m not sure exactly how, but Rackspace gets credit for letting go of OpenStack enough to let the community thrive, for providing one of the more solid distributions out there (Quantum interop aside), for providing a ton of code back to the project and for being one of the more progressive and aggressive OpenStack shops. That’s a hell of a magic trick. Of course, they’ve earned it. Whether it was well designed strategy or good luck, Rackspace stands at the head of the class with OpenStack customers, especially after their announcement targeted at service providers. Also, they easily threw the best party in Portland, so they have that going for them…
Overall, consider me converted. I went expecting a bunch of developers trying hard to be grownups, and I walked away impressed with the depth of talent, scope of ambition, level of organization and amount of energy this community has. If OpenStack wasn’t on your radar before, I suggest you start looking into it. Me? I’ll be learning how to write code.