I’m something of an OpenStack adolescent. Essex was the first release I used heavily. So I’m not a noob, but I’m no graybeard either. This week I unlocked an important achievement in the OpenStack game: I attended Summit.
Since I live near Baltimore, the Juno Summit was a good one to get started with. Atlanta is an easy two hour flight on my favorite airline, Southwest. This trip would get me preferred status on the airline, which is notable since I switched to Southwest in January. (I’ve been traveling way too much.) But this would be worth it.
As it turned out, a number of my colleagues were on the same flight. I sat in the same row as the Internet-famous Nate Burton. However my hopes for engaging conversation were dashed when a human-tank decided to take the seat between us. To make matters worse, the tank watched Gravity on his iPad. I couldn’t look away, and unfortunately it’s not a great choice as far as in-flight movies go, particularly if you have flight anxiety.
After successfully navigating the somewhat-confusing subway system at ATL, the Baltimore OpenStack crew hopped the MARTA light rail to downtown Atlanta. Interestingly, the MARTA escalator there is the longest in the Southeastern US.
Registration was quick and efficient since it was the day before Summit proper. A tip from my colleague, Chris, brought us to Der Biergarten, where I learned that dunkelweizen is better by the liter.
We spotted Red Hat EVP, Paul Cormier, enjoying some beers with Sage Weil, founder and CEO of Inktank. It was a fun evening.
The following morning a sea of Stackers hiked across Atlanta to wind their ways into the bowels of the GWCC, where the keynotes would be held.
The first thing that struck me was how professional everything looked. A top-notch venue, fancy signage, photographers, videographers, everything about the general session said, “OpenStack has grown up.” This project is no longer a geek insurgency. As Troy Toman suggested later, it’s a full-blown Rebel Alliance ready to do battle with proprietary Deathstars.
Jonathan Bryce took the stage to address an audience that was over four thousand strong. Like at previous summits, he said he had planned to invite the code contributors on stage, but there were simply too many. By the current, ninth OpenStack release, codenamed “Icehouse,” the submission process is fairly robust. For contrast, he compared today’s code submission process with how long things would have taken if it were the 19th Century, which he estimated to be about 2 months per submission. Amusingly, I overheard some Stackers say they wished their features would get adopted that quickly. So maybe there’s still some room for improvement.
Jonathan went on to interview Glenn Furgeson of Wells Fargo, and Chris Launey of Disney — both from Fortune 100 companies, both of whom have deployed OpenStack in production ((If you’re interested in more photos of the keynotes, check out my photostream.)). Chris Launey praised the increased speed that OpenStack has brought to Disney. When referencing the axiom, “you can have fast, good, or cheap, pick two,” he said that with all the “fast” you get from OpenStack, you can make your own “good and cheap.” To drive the point home, he equated OpenStack to a “fast-lover’s pizza with an extra helping of fast.”
Troy Toman of Rackspace took the stage next. Rackspace was one of the original contributors to OpenStack release Austin, so I was excited to hear what he would have to say. He did not disappoint. (Although, I admit, numerous Star Wars references, as well as a shout-out to my beloved Sun Microsystems helped.)
Troy described OpenStack as the world’s first “planet-scale” cloud OS. Troy envisions a world with many hybrid clouds, each tweaked for its owners’ particular use-case, but inter-operable with federated identity management. This vision could truly enable the “Internet of Things,” so the term “planet-scale cloud OS” is fitting.
Perhaps the most important thing Troy discussed was the OpenStack DefCore Committee. DefCore is an effort to define the process by which a project can achieve Core status. What are the minimum standards for products with the OpenStack label. What capabilities must an OpenStack product have? What is the code standard? What tests should be “must-pass?” Further, DefCore defines four key attributes for a Core project:
- Shows proven usage
- Aligns with technical direction (translation: jives with the Technical Committee)
- Plays well with others
- Takes a system view
I’m most interested by the second requirement. Every project needs strong technical leadership, particularly open source projects, without which are in danger of the dreaded fork. To OpenStack’s credit, the Technical Committee (TC) meets in the open on IRC. They post minutes to their meetings. A motivated Stacker could review the minutes to learn the thinking of the TC for fun and profit.
In order to shorten an already epic-length post, I’ll only mention the other high-points from Troy’s excellent talk:
- RefStack – The vendor-facing toolset for registration of interop-compliance endpoints and credentials for on-demand testing
- DefCore completion by the Paris Summit (hopefully)
- Not everything can be crammed into the Core, so an Ecosystem of OpenStack products needs to orbit the central components. I’ve been a fan of Project Solum, which Troy mentioned should be able to co-exist with other PaaS if DefCore is successful.
- Rackspace Cloud is expanding its OpenStack-based services. Heat is an example. I’m excited by this; I can’t wait to get my personal stuff off AWS.
- In reference to the inherent tension between open source community members, Troy nailed it with this quote: “Conflict without trust is politics. Conflict with trust is a search for the truth.” This is profound. I encounter this frequently as an employee of an open source company. I’m grateful to have words to describe it.
- OpenStack needs contributions from more than just developers. OpenStack cannot succeed without contributions also from users and operators. If the community can get this right, OpenStack may very well become a planet-scale cloud OS.
Troy’s talk was excellent, and I was thankful for his insight.
Next up was Sam Greenblatt, VP and CTO of Dell. I didn’t know much about Sam before his talk. But after some research, he appears to be more than a hardware vendor executive — back in 2004, Linux World Magazine named him a “Linux Luminary.” Apparently, dude’s got some open source cred, and as somebody who’s been involved with Linux since at least 1994, is a true graybeard.
Sam spoke at length about Dell’s cloud strategy, the details of which can be re-watched on Youtube. You might be thinking, since when does Dell have an OpenStack play? They have some contributions, but they’re pretty minimal ((Dell has made a total of 42 contributions across all versions, according to Stackalytics.)). Where is Dell getting the software chops to do this? Well, they partnered with Red Hat. Tim Yeaton from Red Hat took the stage to talk about how Red Hat and Dell will be enabling the vision of Open Hybrid Cloud.
I was pretty excited about Tim’s announcement of the open source release of ManageIQ, which got more than a few claps. In late 2012, a company called ManageIQ was acquired by Red Hat for a cool $100M, and rebranded as CloudForms. It’s been one of the hidden gems of Red Hat’s product portfolio, possibly less well known due to the fact that it was the company’s only closed-source product. Red Hat open-sources everything — which is one of the reasons why I like working there so much — so this was the moment the company made a $100M donation to the OpenStack community. ManageIQ, now the name of the community version of CloudForms, enables Open Hybrid Cloud management, which sounds complicated, but is actually really cool. The ManageIQ community is still working on security testing, after which the GitHub repo will be unlocked. I was able to leverage my status as a Red Hatter, and get a nightly build of the appliance, and take my word for it, it’s pretty darn cool. From a single interface you can manage OpenStack, AWS EC2, VMware, and RHEV. It’s got a rich, expandable automation framework. And with its “SmartState Analysis” feature, you can monitor and inventory instances even when they’re off or in an error-state — all without an agent or login user. The rest of the world will be able to play with it in about two weeks.
At last the General Session wound to a close, and the Expo floor opened, along with the reason why most people flew all the way to Hot-lanta: the sessions. I was quick to hit up my first selection of the week: Multi-tenant Bare Metal Provisioning with Ironic. Ironic has always seemed to me to be one of the most important projects in incubation. As somebody who once had to design ways to lay down the initial OS images on new systems, I’ve been eagerly awaiting an OpenStack-native way to do it. I discovered that it’s easy to do badly, and you can set yourself up for weeks — if not months — of pain because you didn’t properly build your environment in the first place. And once you add scale into the equation it gets even more difficult.
I was crushed when I heard the speaker open with, “Well I haven’t really done anything to prepare for this, so we’re just going to talk…” You’ve got to be kidding me! Even when I’m just speaking at a meetup, I put days of prep into it. For Summit, people have flown in from all over the world. At the Ceph party (see below) I met a Stacker from Australia, who had flown nearly 20 hours to get to Atlanta. Regardless of how busy you are, there’s no excuse to under-prepare for something as big as Summit. I was even more saddened to discover similarly unprepared speakers throughout the week. This is an opportunity for the OpenStack community to grow-up a bit. Bringing anything short of your A-game to Summit is unacceptable.
I headed down to the Expo floor to see what cool things awaited.
After lunch I booked it up to the third floor for the OpenShift on OpenStack talk. Diane Mueller was there, whom I had met last year in Boston. It was that trip that ultimately brought me to Red Hat, so I was excited to see her again. OpenShift is Red Hat’s Platform as a Service (PaaS) solution that will run anywhere supported by RHEL. RHEL works great in OpenStack, so it’s a great infrastructure for OpenShift. To make deployment easier, OpenShift publishes Heat templates on GitHub. There’s a demo on Youtube similar to what Chris Alfonso presented in the session. Chris tried to think ahead by showing a fast-forwarded video of a Heat deployment, thereby relieving the audience of lengthy waits, but slow Summit wifi smote the demo. Lesson learned: always a) run locally or b) bring your own WAP.
After numerous sessions and hallway meetings, a few of us hopped a cab to Buckhead to go to the most excellent Umi. It’s been on my sushi-aficionado bucket list, right up there with Sukiyabashi Jiro. (With any luck a future Summit will be scheduled for Toyko and I can check that off my bucket list.)
Not usually a fan of Japanese whiskey, the Yamazaki 12 year went nicely with the wide variation of flavors of the evening. In scotch terms, it reminded me of a Highland malt, very light and dry.
A number of us had intended to go to the Piston party, “A night at the Opera, at the Opera.” But after the 10-12 courses of sushi, we were late, tired, and full. Better to turn in early and get a head start on the next day of sessions.
The plan must have been sound, because Mark Collier, the first speaker of Tuesday’s general session, who appears to be quite the fan of Breaking Bad, made several references to late-night Piston shenanigans. More than a few attendees looked a little worse for wear. Mark had several interesting guest speakers, including Guillaume Aubuchon, CTO of DigitalFilm Tree, and Toby Ford, AVP of AT&T.
After Mark’s other interviews, the folks at Solidfire took the stage. They’re another hardware vendor offering another OpenStack-in-a-box product. But what I found interesting about them was the fact that, in partnership with Red Hat and Dell, they’re publishing their OpenStack reference architecture. I find this to be pretty useful. Even if you don’t plan on deploying a flash-based architecture, it’s great to have a starting point — especially if you’re dealing with customers who have never deployed OpenStack before.
The final speaker of the general sessions, whom I have to admit really nailed it, was Mark Shuttleworth. Mark has a long history with Linux — Debian, in particular. In October 2004, Mark assembled a team of Debian developers, and released the first version of Ubuntu, 4.10, codenamed “Warty Warthog.” I used to be an Ubuntu user, myself. It’s quite popular, and I wanted to see what all the fuss was about. But I’ve always used rpm-based distros — as far back as Red Hat Linux 5.2 — and I found SELinux to make Ubuntu unstable. Maybe it was operator-error, but Fedora is so excellent now, that I made the switch back. In any case, Mark’s keynote was engaging and funny. You can watch it here. I give extra points for doing a live demo on stage — always risky, so I tip my hat to that.
After Mark’s keynote I went back to the Expo floor. The day before I had noticed a number of booths posted a flyer saying, “Red Hat OpenStack Certified Partner.” Out of curiosity I wanted to see how many of the booths were flying Red Hat colors. It seemed to me the best way to remember would be to take photos, so here they are. I’ve also made a map of booths with partner status. For data freaks out there, that was 36 out of 90, or 40%.
One thing we kept hearing in the keynotes was how important the OpenStack Ecosystem will be. Well, if you’re a Red Hat user, it looks like the Ecosystem is off to a great start.
Over the next three days I attended a number of sessions, hung out in the dev lounge, and attended (probably too many) parties. Photos of shenanigans follow.
It was a fantastic Summit, and I was honored to attend. Maybe next time I’ll have something useful to contribute.
Can’t wait until Paris!