AWS 2014 Trip Report - From the Dev's POV

Tara Calihman - April 02, 2014

This was my first time working a conference, period. I’d been to a couple as an attendee, but not from the vendor side. It was educational, both from the perspective of a future attendee (be nice, have actual questions, if you’re not interested, don’t pretend to be), and from a vendor perspective (the importance of swag, how simple mistakes of confusion in the demo can mount together very quickly, etc.).


One thing in particular that grabbed me was a reality check in terms of being known. I knew we weren’t big fish, but getting out of Boulder, I had a bit of a revelation in that no one knew who we were. That seems to me a really good thing honestly - I had been somewhat afraid that we’ll be having to do a bunch of work to undo perceptions from our beta showing (which was good, but still very beta), but it turns out that the pre-conceptions I was worried about just don’t exist.

This space is going to be super important. No one questioned what we were doing, and, after explaining the vision, literally no one needed help understanding why we’re trying to solve for what we are. We had a couple people who thought the problem has already been solved, or that particular parts of our implementation aren’t the right way to solve the problem, but everyone I talked to understood that the problem exists.

My time there also reinforced the importance of the whole monitoring/alerting/collaboration trifecta. A huge number of the vendors there do monitoring of some sort (Splunk, DataDog, etc.), and everyone we talked to had some sort of alerting system in place (whether it was our friends at PD, or some homegrown solution around google groups, etc). But once the alert is delivered to the user, no one had any process in place more robust than “hit IRC and hope the people you need are online.”

Given how I think VictorOps can solve that last part completely, I’m feeling pretty fantastic about where we’re going. I’m super excited to get a couple of the people we talked to into a room to talk about how we might design some of the features that are coming up soon, and I’m feeling really good about the progress we’ve made so far. We’re hitting our stride on the alerting side, and the roadmap once we get that stuff figured out is going to be really, really fun.

Other observations from the conference:

  • “Cloud” is an awful marketing term. I can write a screed on this later, but a server is a server is a server. A ridiculous number of people asked me why we were at an AWS event if we’re not AWS-based. Uhm, what?

  • Remediation is going to be powerful. A bunch of people asked if we could do anything like auto-bouncing servers on failures or spinning up extra EC2 instances, etc., and then immediately recanted when I suggested that we’d prefer to stay off your servers explicitly, and they thought through the security implications of that. That said, I can see a ton of ways that we could make the current “web hook as escalation” remediation concept work better for doing exactly that sort of thing.

  • There is definitely something  to be said about drinking at sea-level.