AWS re:Invent is the annual “all things AWS” conference, with everything from sessions to learn more about existing services, the announcement of new features or entirely new services, to the formation of a new league for racing remote control cars powered by machine learning. It’s held in the desert, by which I, of course, mean Las Vegas. This year is the first time I’ve personally attended, so this is my newbie trip report along with plenty of free advice that I may or may not have actually followed. It bears repeating that if you happen to get sucked in by craps tables or the various other vices that Vegas incessantly pushes on you, don’t go to this conference, because it’s inescapable. Public service announcement over.
It’s easy to underestimate everything that’s involved in planning a feasible schedule for yourself over the course of a week in Vegas. Several aspects of the planning process are more challenging than they appear at first, especially once the scale of the event became apparent to this newbie here. First of all, there are at least 5 different venues hosting sessions all up and down the strip for 50 thousand attendees. There are several different types of sessions, from the traditional “sit and listen to a lecture”, to a longer and more involved workshop-style training where you’re building something. Other types of sessions are mostly Q&A after a short intro, so people can talk through questions with experts in a particular area.
Pre-registration for sessions opens up about 6 weeks before conference time, and reserved spots for popular sessions fill up fast. It’s a good idea to read through the session catalog before pre-registration and mark the ones you’re interested in so you’re not starting from scratch while others are racing to reserve a spot in their favorite talks. There are some talks that aren’t available for pre-registration because they’re about new services or features that are announced during the conference in the keynotes, so keep an ear out. They’ll get added to the catalog as they’re announced, so if a new service pops up that piques your interest, fire up that conference app on your phone and go for it!
Logistics come into play already when considering what you’re interested in, for example, an hour is recommended to get from one hotel to another, so you can’t realistically schedule talks right after each other. And you have to eat too. There are “to go” boxed lunches available, or you can sit down in the dining area for better-tasting hot food. The first day, it was a happy accident that I ate lunch right as the dining hall opened because it got crowded quickly. I did my best to eat early, schedule permitting, for the rest of the week too. The desserts were surprisingly good, my favorite was the mochi. Oh, and find the closest Walgreens to where you’re staying, it’s a life-saver whether you need a toothbrush, a snack because you skipped lunch to go to a session, or you just don’t feel like paying quite as much for a beer.
With regard to reserved capacity, anecdotally at least, this year was handled better than last year. I heard stories of people at last year’s re:Invent who reserved seats and didn’t get into the talk because it had already filled up. 2018 was apparently much better, in that the number of reserved slots was lower than the total number of seats available, so all the reserved people got in, plus people from the “walk up” line. Lines start forming for both reserved and “walk up” an hour ahead of session start time, so with the number of venues and all the waiting in line, it’s entirely possible to spend more time transitioning between sessions than in the actual sessions if you’re not careful. And oh, the lines. Lines everywhere. Lines at registration, to pick up swag, for lunch, for shuttles, for sessions, to leave sessions, for games at re:Play, and for a dozen other things.
With all that said, I read a lot of advice, and still managed to end up with a schedule that had me in 4 different casinos on the first day, chasing interesting topics up and down the strip. Here’s how that went! We stayed in the Venetian (that’s 1). My first session was a mile South in the Aria (now we’re at 2) at 10 AM. I walked. It wasn’t that bad. Lunch is served in each venue starting at 11 AM. My next session was in the MGM Grand (up to 3) at 1:45 PM “just across the street”, but it’s farther than it looks. My last session of the day was in the Mirage, which would have been 4 (nope) but I was tired of walking by that point. We also had a booth to staff, and the first night was the welcome party in the expo with food, beer, and all sorts of free stuff covered in logos. I had less travel in my schedule the rest of the week, mostly staying in one location all day, but wow, that first day I was all over the place.
So with all that bouncing around, what did I learn? Well, I learned that the battery in my old laptop can’t go all day anymore, so half my notes ended up on my phone. All (well ok most) joking aside, I went to several very good talks. I’ll summarize a few here, with the tiebreaker being the order I saw them in.
First place was kind of not fair. All sorts of new features, services, and, um, cars were announced at the various keynotes. Some of the re:Invent advice I read recommended attending at least one keynote in person, so I went to Werner Vogels’ on Thursday. Watching the DJ do his thing up on stage in a dark room full of thousands of people at 8 o’clock in the morning was pretty entertaining too.
Next was the first talk I went to, CON310 “Breaking Containers: Chaos Engineering for Modern Applications on AWS”. I picked this talk because here at SparkPost, more of our services are running in containers every quarter, so it seemed relevant to start thinking about how to inject certain types of failure. For those of you familiar with the phrase “error budgeting“, this kind of poking and prodding is a great way to spend that budget and end up with useful information about your system in exchange. I enjoyed this talk because there were references to other sources (books, videos, etc) to get more in-depth information, and also because it raised questions to ask about your infrastructure and talked about ways of thinking more than just the details of one way to solve a specific problem. Well worth a watch.
Another good one was SRV373 “Building Massively Parallel Event-Driven Architectures”. This talk was a good overview of the various “serverless” services offered by AWS. A MapBox engineer also described a neat tool to help with “Lambda-backed custom CloudFormation resources”. I picked this talk because more and more parts of our system are using serverless technology, in a variety of ways, and I enjoyed hearing about gotchas and use cases from people who build and use the underlying systems every day.
There were lots of other good ones, and I had a great time not just learning, but interacting with people in sessions, at lunch, and at our booth: customers, people with questions, and the rest of the crowd that wanted our free stuff. In summary, it was a great experience, and I even learned some stuff. Plan ahead, you can’t do everything, and see you next year?