AWS re:Invent 2019 with Corey Quinn: The biggest announcements from the conference

TechRepublic’s James Sanders spoke with Corey Quinn, cloud economist at The Duckbill Group, at AWS re:Invent. Quinn has developed a reputation for hot takes on AWS service offerings and billing practices, and his outspoken nature has made him a regular fixture on Twitter. The following is an edited transcript of the conversation. 

Corey Quinn: From my perspective, one of the most interesting things that has come out based upon what I do is Compute Optimizer. It uses machine learning and instead of applying it to ephemeral things that can find anything in a pile of data except a business model, it applies it to a bounded problem domain of the AWS bill. Are you running instances that are properly sized? And if not, here’s what we recommend instead. It’s valuable, and it’s helpful for companies that are trying to dial in their workloads.

Graviton2 is interesting, and I think we’re going to see the rise of ARM in the somewhat near future. People are going to do some analysis and figure out what’s it take to port our workload from x86 over to ARM, and the answer is just move it–it already works. It also turns out that a little more work needs to be done. Cross compiling isn’t just about using GCC when you’re pissed off.

SEE: AWS Graviton2: What it means for Arm in the data center, cloud, enterprise, AWS (ZDNet)

CPU performance and cost savings

The interesting part is that for similar or reportedly superior performance, they’re seeing roughly 20% cost savings. Don’t quote me on that. I don’t memorize facts and figures the way some folks do. Now, the counter argument is … so at the point where they start having that level of access to discounting, companies are going to start to make the shift on their own. It feels like we are on the verge of a revolution towards ARM, and what’s going to probably start tipping it over is if Apple releases a laptop with an ARM chip in it, and honestly, the fault of this largely rests with Intel. A bunch of blown deadlines, a bunch of overpromising and underdelivering has really led to a point where a lot of manufacturers seem fairly disgusted with them. Now, Intel has a booth here at the re:Invent Expo Hall, but I can’t for the life of me figure out why.

SEE: Intel’s horrible, bad, terrible week amid AWS, Qualcomm’s Arm moves (ZDNet)

That’s what’s going to become so fascinating, as we see future generations of instances coming online across all the different processor manufacturer families. What are we going to see performance wise? And not for nothing, as much as we all love to talk about how critical CPU performance is, you take a look at any traditional workload in most businesses, and the computers are incredibly bored. It’s like watching an IBM keynote.

AWS Outpost and hardware vendors

James Sanders: What impact will AWS Outpost have on traditional hardware vendors, like HPE and Dell?

Corey Quinn: I don’t necessarily know how to answer that question because I don’t tend to deal with companies with massive on-premises environments just based upon the nature of what I do, or if they have them, they’re rapidly looking to evacuate them for a variety of reasons. I don’t think that those companies are going to dry up and blow away anytime soon, but I think that there is a definite longterm story for them. If you’re starting a Fortune 500 today that’s going to grow from small, scrappy startup into that giant company, where’s your on-ramp for some of the traditional enterprise IT vendors? The unfortunate truth is, as things stand today, you’re never going to become their customer. They need to find a way to bridge that gap in a compelling and authentic way, and right now I’m not hearing a terrific roadmap to get there being articulated from any major player in this space.

Amazon Braket is quantum in the cloud

Amazon Braket, or however you want to pronounce it, or briquettes? Pronunciation is the most interesting thing about it because right now, there’s a whole lot of talk about it, and absolutely nothing demonstrative to show about it. It is future technology that has no real-world presence or application today in any meaningful sense. When that changes, they want to be able to say they’ve had something and they’ve been looking at it for a long time, but there is nothing as a paying customer I can buy today in the world of quantum that solves a problem I have. And until that changes, we are fundamentally not going to see significant traction in that market. But given that their competitors are saying something, they have to as well.

SEE: Amazon Braket quantum computing service glues together three unrelated systems (TechRepublic)

Amazon’s new products named after women

Generally speaking, we take a look at things like Amazon Macie and Kendra that was just released. Generally speaking, if Amazon names a product after a woman, double check the pricing because it is likely going to be extortionate. I do not know or understand why they have chosen to go down that path; Kendra is awesome, but it is incredibly expensive. Macie is interesting, but laughably expensive. Again, they have no monopoly on being expensive and confusing from a pricing model perspective, but it is interesting to see them charge on new and exciting dimensions.

CodeGuru is also too expensive

The Amazon CodeGuru, which I feel like A Cloud Guru‘s response is, “I’m sorry. You called it what now?” It’s fascinating. It charges 75 cents per month per 100 lines of code review. First, just stop including line breaks and use a crap ton of semicolons. Secondly, it’s a terrible metric for developers, specifically. And thirdly, even if the pricing is ultimately reasonable, suddenly people are having to think about their code and their development practices in a way that until now they haven’t, and that becomes a significant problem. I think it is a poor choice of pricing model. I’m interested to see how well it works, but today it is laughably expensive. I have applications running with the serverless framework written in JavaScript or Python that cost pennies a month that would cost hundreds of dollars to put through this service.

But Wavelength has potential

Amazon Wavelength is a genius product, and I say this sincerely, because 5G is another marketing buzzword where everyone is going to have to play in this space. For better or worse, we do see wireless spectrum is relatively commoditized at this point. Some providers are slightly better than others, but there’s no concern of which provider I have dictating whether I can make a phone call or hit the internet or not.

SEE: AWS re:Invent 2019: Why Outposts, Local Zones, and Wavelength are game changing for enterprises (TechRepublic)

5G is a buzzword everyone’s targeting. But what makes this fascinating from my perspective is by positioning themselves in partnership with Verizon, Amazon is getting in on the ground floor of 5G development, which grants them access to what’s going on in that space, and their investment level really distills down to a bunch of racks of hardware they’re slamming into a point of presence that Verizon owns. It’s a great opportunity, from Amazon’s perspective.

From Verizon’s perspective, they have the same problem that most telcos seem to, which is the world wants them to be a dumb pipe, but they can only charge so much for that. Well, most companies. Verizon has fascinating ways of getting around that. Verizon Math is still a thing, but what’s neat there is that they want to start offering higher level differentiated services, but no one wants a service that Verizon runs other than the connection itself. So, we’ll see if it turns into a competitive differentiator or not. Time will tell, and I will make fun of it either way.

I will say that Verizon is almost certainly shocked and astounded by how positively the community is responding to Wavelength. It is the first time in history that Verizon has gotten a good reception.

SEE: AWS, Verizon aim to collaborate on 5G, cloud, edge computing use cases (ZDNet)

Jassy’s keynote speech

Andy Jassy gets onstage and has an almost insurmountable problem that he has to solve for. He asked to articulate a huge variety of service releases and enhancements to an incredibly diverse audience who is coming from a huge variety of industries, backgrounds, and problem spaces. Every service release is important to someone, and no service release is important to everyone.

Having to strike so many different notes to so many different constituencies about so many different things as well as doing a marathon three-hour keynote session is incredibly challenging, and I have to say for better or worse, he is more polished every year as a speaker. As someone who gets onstage a fair bit myself, mostly due to my ongoing love affair with the sound of my own voice, I know what it takes to speak in front of a lot of people. He speaks to way more than I generally tend to.

The AWS house band needs better lyrics

Oh yes, the house band is depressing, sad, and honestly, if Amazon were a kinder, warmer company that expressed empathy better than they do, they would put the re:Invent house band to sleep. The problem that they’re having is that it’s loud; it’s very off. They do song covers very briefly, but they don’t change the lyrics to be anything that would be humorous or inspire people to look at things differently, so it just feels a little hokey. Give me 20 minutes with the lyrics and let’s see what we can come up with instead. If you’re going to go for something, go all in.

I think there are always going to be surprises. Amazon is a company that doesn’t hold still very long or very well, it’s a trait that I tend to share. The difference really comes down to the fact that I’m good at naming things.

Also see