Amazon Web Services (AWS) just wrapped up re:Invent 2016--their annual week of education, networking, announcements, etc. This year's conference was their largest ever, including around 25,000 attendees and more vendors than I had time to visit. Each year, this event seems to set the tone for the next year of AWS in terms of direction and emphasis. This year was no exception.
In this post, I will explain the main trends and takeaways that I got out of the experience.
Serverless is really taking hold
If there was one key trend that I observed at the conference, it was that interest in serverless architectures and technologies is enormous. Every session I attended on the topic was at capacity, often with additional people waiting in line for a chance to get in. Not only is interest high, but most of the large presentations included case studies of large clients who are already running significant serverless applications in production. Additional announcements from the conference should help close some gaps that may have prevented users from adopting Lambda. Clearly, serverless has everyone's attention.
I'll go a step further here and say that I won't be surprised if serverless overtakes containers as the preferred method of application design and hosting. Serverless is easier to use, and avoids several of the challenges that come with containers. However, it does have the disadvantage of increasing vendor lock in (more on that below).
Machine Learning is becoming the new norm for analytics
Just like with serverless, the Machine Learning sessions and mini-con were frequently at capacity. The client presentations gave compelling use cases for ML, such as real estate market analysis, predicting the recurrence of patient visits for insurance companies, etc.
What really stood out, though, was how all these companies are using ML to do near real time, predictive analytics. Simply being able to analyze data to gain business intelligence is not good enough anymore--companies want intelligent, instantaneous answers to questions, and they want this integrated with their applications. The idea is that you should be able to predict a customer's next move on a website and respond accordingly; or, decide on the fly whether or not to authorize an insurance sale. The ML APIs support this functionality, and I think that it will be commonplace for applications in the next few years.
Voice Interaction will play a key role in user interface
If you watched any of the keynote, you saw that AWS has opened up the power of Alexa's voice interaction system with the release of "Lex" and "Polly." This new set of features will allow developers to build applications that use voice interaction as a key part of the user interface. Users will be able to carry on a conversation with applications, and issue commands that result in actions. If you own any Alexa device, you have already experienced some version of these interactions. However, with Lex, you will start talking to your phone the way you talk to Alexa, much like "Ok, Google" or Siri. The promise of this suite of technology is that more developers will be able to harness this power to allow full conversations with natural language recognition. It's easy to imagine the release of another device in the future, some sort of wearable Alexa device, that lets you do most of what you do with your phone without touching anything. Personally, I'm hoping for cuff links that I can use to call for Uber--not quite exactly like James Bond, but getting there. My guess is that this type of interaction is still about two years out.
AWS is answering Azure Hybrid with VMWare
One of the key differentiators for Microsoft's Azure platform has been its focus on seamless hybrid cloud solutions. Many businesses gravitate toward Azure for this very reason. I expected AWS to announce new features/functionality related to their on-prem connectivity options in order to answer the challenge. Instead, all they did was reinforce the relationship between AWS and VMWare, which CEO Andy Jassy lauded in his keynote as a simple way to reuse the solutions you developed on-prem in the cloud. Personally, I don't think that is enough to sway customers who are set on a hybrid model, but perhaps AWS sees otherwise, or possibly AWS is discounting the importance of hybrid as a short-term play.
AWS is trending towards Managed Services, and what this means for Vendor Lock-in
It is becoming clear now that AWS is starting to offer more managed services than they have before. In the past, AWS preferred to focus on the cloud infrastructure and their cloud tools, and leave it to the customer to figure out how to use them. Now, several of the new offerings include fully managed services. For example, Lightsail is intended to provide a full environment setup with a single click, and Shield, a DDOS protection service, may be the first time that AWS has crossed their own shared security boundary to manage security for the client. Additional offerings include managing VMs down to the patching level.
With this level of service, it's no wonder that many of my conversations involved concerns about vendor lock-in. Vendor lock-in has been a trending concern over the last year especially, with companies expressing reservations about tightly coupling themselves to any single cloud vendor. Obviously, the more a company utilizes managed services, the more they are locked in to that vendor. There is not clear path or consensus here yet, but I definitely noticed that more people I talked to were taking the path of treating vendor lock-in as an acceptable risk.
Not everyone is in the Cloud yet
If you stay on top of cloud news, it can seem like this whole "cloud" thing has already happened, and everyone is already doing it. That is not the case. I attended several sessions focused on cloud strategy and cloud migration, and these were packed with people who are still figuring this whole cloud thing out. Additionally, one could not walk through the vendor exhibition area without seeing multiple vendors offering planning and migration solutions.
I think that it's more accurate to say that a majority of companies have dabbled in the cloud, as opposed to saying that a majority are using the cloud. The key metric here should be whether or not a company is running a meaningful workload in the cloud already. I suspect that a survey with that metric in mind would paint a picture where 60% of companies are still figuring out how they are going to get to the cloud. But, unlike past years, cloud adoption does seem inevitable now.
As always, it remains to be seen what will truly come of all the announcements made this year. As with years past, some new features will just fail to catch on and eventually disappear. Others will thrive. Remember that Lambda was announced at last year's conference, and while it has seen significant adoption, that growth will likely continue in the future. Perhaps this time next year we will be looking at "Lex" and "Polly" the same way. Regardless of what does and does not catch on, we will definitely be talking about the cloud again.