Serverless conference 2017 tl;da (Too long didn’t attend)

Tim Wagner at Serverless conference 2017

A few weeks ago I attended the Serverless conference in sunny Austin, Texas. Since then I’ve read a few good summaries and want to highlight some areas that I think need to be addressed.

tl;dr Serverless is one of the fastest growing areas in tech right now. Attending the conference I was hoping to be introduced to innovative new applications and use cases that I had never seen/heard about before. The handful of technical talks that I attended were pretty great, but the vendor talks felt much too pushy.

For a summary of specific talks please click here.

Serverless conference is by no means a massive stadium filling affair. Compared to other well known conferences such as Strata-Hadoop the attendance is relatively small (~400 people). This greatly influences the type of content expected by the attendees. Generally speaking, smaller conferences tend to attract early adopters and genuine enthusiasts. As a byproduct of this, the content is usually expected to be far more technical than what is given to the general public. Choosing Austin as the location only perpetuated this expectation. Austin is a great city but it’s not one of the most tech dense cities in the U.S. A handful of the larger tech companies have R&D groups in the area, but this pales in comparison to the likes of the Bay area, New York or Seattle. Therefore most attendees that were not native Texans probably really wanted to be at the conference. Hopefully this will shed some light on the following summary and my frustrations associated with the conference.

First and foremost, one thing is abundantly clearly — the big players have their eyes locked on Serverless. If anyone was doubtful before let me quell those doubts right now. Amazon, Microsoft and Google showed up in force, not only in attendance but primarily in content. Over the duration of the conference I attended at least six talks that exclusively focused on Azure features or functionality given by Microsoft employees or close partners. From my point of view it turned what could have been an insightful and enlightening look into the Serverless world into more of a long-running and repetitive advertisement. This being said, a couple of their talks did present the audience with interesting new features within their ecosystem and at least showed that AWS isn’t a uncontested challenger in the Serverless arena. Unfortunately, Microsoft has a very steep uphill battle, even if Azure Functions were many orders of magnitude better than AWS Lambda there would be few adopters due to the AWS ecosystem lock in.

Although Microsoft by far had the most presentations that came off as advertising, the other players spent a significant portion of time selling their own product. With the exception of a talk by IBM almost all of the presentations by bigger companies could essentially be described as “a live version of the online documentation presented by someone who most likely has never actually used the thing they’re talking about”. Although the information in these talks is still potentially valuable it doesn’t help building credibility when presenting to a large group of developers. This was incredibly disappointing, since from watching the past conference (albeit online) I was given the impression that the the aim was for a more grassroots community driven vibe.

To clarify, I have no issues with big companies attending and even presenting. I’m not disillusioned about the tech space and understand that these companies have to adopt new technologies quickly or risk obsolescence. That being said Serverless has near limitless avenues to explore, and as fortune 100 companies it shouldn’t be difficult to allocate more resources towards coming up with something interesting to discuss for 30 minutes.

In my eyes, a great example of a presentation done right was a talk by the development team working at Nordstrom. These are things I think they did right

A real development team did the presentation

This was the biggest differentiator with Nordstrom’s presentation. They sent up a team of real developers that actually wrote the software that they were demoing. Sure, they weren’t charismatic and charming in the same way that AWS and Microsoft presenters were, but they were people that the audience could relate to. Furthermore, they sincerely seemed proud of what they had done and would often get slightly off track in the way you would expect someone to when they believe in something.

It wasn’t a sales pitch

Clearly the type of products Nordstrom’s sells is not as applicable to the conference (in comparison to the likes of Google and Microsoft) but I think the point is moot. The team took the stage with the full intention of showing a cool piece of technology that they had developed for themselves. I didn’t feel like they were trying to convince or hook me but rather that they were presenting facts and opinions (which were strongly stated as such) and leaving me to draw my own conclusion or meaning from them. Without a doubt this is the most important quality that so many of the other presenters failed to adequately achieve.

It wasn’t platform specific

I believe they mentioned that they ran on Lambda (which is absolutely acceptable) but they didn’t try and make Lambda out to be the second coming. They were honest and highlighted both advantages and disadvantages of their experience with the product while keeping the discussion about it brief. This is key in today’s world of corporate influence, since it’s often hard for consumers to discern whether they are being played or not.

They sent the best team they could

The team that gave the presentation had actually won an award whose prize included a trip/talk at Serverless conference. Everyone on the team seemed genuinely proud of this award, which to me gives whomever is presenting a lot more credibility.

My theory is that most of the speakers went with the “live billboard” approach out of desperation. At last year’s conference AWS Lambda was for all practical purposes the only viable option if you were considering Serverless development. Sure, Azure had an offering but it was in its infancy and relatively under-represented. Furthermore, Google hadn’t even made its first real step into the Serverless world. This caused many of Lambda competitors to go into a frenzied panic, believing that AWS had a dangerous head start in the space. This manifested at the conference as a bunch of companies trying to convince developers of why they shouldn’t be using Lambda, as opposed to why they should be using Serverless.

The bottom line is that developers are generally resourceful and independent thinkers. If Azure is offering functionality that makes Lambda a less viable option developers will find out about it (conference or no conference). This mindset definitely affected the atmosphere of the conference.

Big players aside, there were a large portion of presentations given by smaller companies or independent developers. These were primarily centered around the issues managing or developing Serverless applications and tactics to mitigate them. With a few exceptions, almost every presenter spent a large portion of their time attempting to sell their product. This may seem logical at first but the presenter guidelines specifically and directly disallowed this. Why these talks all made it through the vetting process may be indicative sign by itself, but there isn’t much value to be had dwelling on it now. Instead I’ll discuss what I think a few of the presenters did absolutely right:

Charity Majors | Serverless: the early years

By far my favorite presenter. She didn’t try and sell me anything, was direct, honest and completely uncensored. Adding to this she didn’t merely describe her solution to a pre-existing, well defined deficit with Serverless technology but instead questioned many of its assumed foundations. In addition, she even criticized many of the problems I outlined regarding bigger companies during her talk. Due to this, she was one of the most well received and discussed presenters of the conference. Charity’s approach captures exactly what developers are looking for at these conferences, they want to be told why what they’re currently doing/thinking is wrong (and not even necessarily how to fix it!). She was one of a few presenters to involve the audience and it paid off in dividends.

Ben Kehoe | What’s Missing From Serverless Providers

Ben discussed issues and pain points that his company (iRobot) had encountered when migrating to a Serverless approach. Specifically, he stressed his belief that, because Serverless thinking is so inline with a microservice methodology, there needs to be a greater focus on providing support to connect the two paradigms. I think his ideology is very inline with things I’ve thought about and discussed personally, so the content was immediately relatable.

Guy Podjarny | Serverless Security: What’s Left To Protect?

Guy, too, treaded a line between sales pitch and informative presentation, but he managed to do it very well. I valued his attempts to find unforeseen issues within the Serverless ecosystem, which is something that doesn’t happen enough. His discussion was strongly grounded in reality and in many ways it popped the euphoric Serverless bubble that has been created by Amazon and others over the last two years. His overall message could be summed up in a single statement, “Attackers are developers; when technology progresses so do they”.

Keith Horwood | Service Composition with StdLib

Many of the concepts and ideas contained in Stdlib are ones we (at Binaris) are discussing as well. Stdlib is not a new offering but in recent months they have added more and more depth to their stack. As of today Stdlib not only defines the format of your functions but also obscures the underlying execution, therefore locking you strongly into their ecosystem. I personally have qualms with this type of play since it leans much more towards the restrictive Microsoft model of user interaction. To make it clear Stdlib doesn’t provide a custom run-time but instead simply wraps Lambda and then makes it impossible for the user to access the underlying call. This means you lose all convenient Lambda integrations such as built in event sources which is one of the main draws of Lambda in the first place. This is unfortunate to see happening since many of the higher level API ideas from Stdlib are right on track.

Conclusion

Overall I was disappointed in the severe commercialization of the conference. It seemed to move directly away from the very popular and successful developer oriented approach of the previous conferences. This is so strange to me because history has made it abundantly clear that you cannot sell to developers like you would traditional consumers. There were some nice takeaways and from an organizational standpoint the conference went off without a hitch but in the end it fell short. It was incredibly valuable being able to interact and communicate with so many developers interested and using Serverless. With that in mind I hope and believe that other attendees felt similarly to me and they revert to the previous formula in future conferences.